09-08 Accelerated Performance Testing On The 2006 NCAT Pavement Test Track

ACCELERATED PERFORMANCE TESTING ON
THE 2006 NCAT TEST TRACK
By
Richard Willis
David Timm
Randy West
Buzz Powell
Mary Robbins
Adam Taylor
Andre Smit
Nam Tran
Michael Heitzman
Alessandra Bianchini
December 2009
ACCELERATED PERFORMANCE TESTING ON THE 2006

NCAT TEST TRACK
By
Richard Willis
David Timm
Randy West
Buzz Powell
Mary Robbins
Adam Taylor
Andre Smit
Nam Tran
Michael Heitzman
Alessandra Bianchini
National Center for Asphalt Technology

Auburn University, Auburn, Alabama
Sponsored by
Alabama Department of Transportation
Florida Department of Transportation
Georgia Department of Transportation
Indiana Department of Transportation
Mississippi Department of Transportation
Missouri Department of Transportation
North Carolina Department of Transportation
Oklahoma Department of Transportation
South Carolina Department of Transportation
Tennessee Department of Transportation
Texas Department of Transportation,
Federal Highway Administration
Oldcastle Materials Group
December 2009
Willis, Timm, West, Powell, Robbins, Taylor, Smit, Tran, Heitzman and Bianchini
ACKNOWLEDGEMENTS
This project was sponsored by Alabama Department of Transportation (DOT), Florida DOT,
Georgia DOT, Mississippi DOT, Missouri DOT, North Carolina DOT, Oklahoma DOT,
Tennessee DOT, Texas DOT, Federal Highway Administration, and the Old Castle Materials
Group. The project team appreciates and thanks Alabama DOT, Florida DOT, Georgia DOT,
Mississippi DOT, Missouri DOT, North Carolina DOT, Oklahoma DOT, Tennessee DOT,
Texas DOT, Federal Highway Administration, and the Oldcastle Materials Group for their
sponsorship of this project.
DISCLAIMER
The contents of this report reflect the views of the authors who are responsible for the facts
and accuracy of the data presented herein. The contents do not necessarily reflect the official
views or policies of Alabama DOT, Florida DOT, Georgia DOT, Mississippi DOT, Missouri
DOT, North Carolina DOT, Oklahoma DOT, Tennessee DOT, Texas DOT, Federal Highway
Administration, the Oldcastle Materials Group or the National Center for Asphalt
Technology, or Auburn University. This report does not constitute a standard, specification,
or regulation. Comments contained in this paper related to specific testing equipment and
materials should not be considered an endorsement of any commercial product or service; no
such endorsement is intended or implied.
ii
TABLE OF CONTENTS
LIST OF ACRONYMS ............................................................................................................ 1

EXECUTIVE SUMMARY ...................................................................................................... 3
CHAPTER 1 - INTRODUCTION ............................................................................................ 6
CHAPTER 2 - OVERVIEW OF THE 2006 TEST TRACK (PHASE III) .............................. 8
EXPERIMENTAL DESIGN .................................................................................................... 9
CONSTRUCTION .................................................................................................................. 12
TRAFFIC ................................................................................................................................ 14
PERFORMANCE MONITORING ........................................................................................ 14
CHAPTER 3 - MIXTURE PERFORMANCE COMPARISONS .......................................... 16
VALIDATION OF THE ENERGY RATIO CONCEPT ....................................................... 16
Laboratory Energy Ratio Determination ............................................................................ 17
Observed and Mapped Cracks ............................................................................................ 17
In-Situ Pavement Response ................................................................................................ 18
Core Observations ............................................................................................................... 18
Bond Strength Testing ........................................................................................................ 18
Energy Ratio Conclusions................................................................................................... 19
LABORATORY AND APT TESTING OF MODERATE AND HIGH RAP CONTENT
MIXES .................................................................................................................................... 20
Field Performance ............................................................................................................... 21
Laboratory Testing .............................................................................................................. 22
Summary ............................................................................................................................. 24
LOW NOISE PAVEMENT STRUCTURES ......................................................................... 25
Low Noise Pavement Design .............................................................................................. 25
Materials ......................................................................................................................... 25
Surface Macrotexture Measurements ............................................................................. 25
Sound Testing ..................................................................................................................... 26
Sound Pressure and Intensity Measurements ...................................................................... 26
Findings............................................................................................................................... 28
Macrotexture ................................................................................................................... 28
Sound Pressure ............................................................................................................... 28
Sound Intensity ................................................................................................................ 29
Sound Absorption ............................................................................................................ 29
Summary ............................................................................................................................. 29
TEST TRACK FRICTION ..................................................................................................... 30
CONSTRUCTION AND PERFORMANCE OF PERMEABLE SURFACE MIXES .......... 32
Experimental Design ........................................................................................................... 32
Placement ............................................................................................................................ 33
Laboratory Performance ..................................................................................................... 34
Field Performance ............................................................................................................... 34
Noise ............................................................................................................................... 37
Drainage ......................................................................................................................... 42
Conclusions and Recommendations ................................................................................... 43
iii
DETERMINATION OF RUT DEPTH CRITERIA FOR LABORATORY PERMANENT
DEFORMANTION EVALUATION...................................................................................... 43
Experimental Plan ............................................................................................................... 44
Specimen Preparation ......................................................................................................... 44
Test Results and Analysis ................................................................................................... 45
Conclusion .......................................................................................................................... 50
LOW AIR VOIDS EXPERIMENT ........................................................................................ 51
CHAPTER 4 - STRUCTURAL STUDY ............................................................................... 54
EXPERIMENTAL OBJECTIVES ......................................................................................... 54
2006 TEST SECTIONS .......................................................................................................... 55
Florida DOT: Sections N1 and N2...................................................................................... 55
Alabama DOT and FHWA: Sections N3 N7 ................................................................... 55
Oklahoma DOT: Sections N8 and N9 ................................................................................ 56
FHWA: Supplemental N9 ................................................................................................... 56
Missouri: N10 ..................................................................................................................... 57
Alabama DOT: S11............................................................................................................. 57
INSTRUMENTATION .......................................................................................................... 58
CHAPTER 5 - MECHANISTIC PAVEMENT ANALYSIS ................................................. 59
TEMPERATURE AND SPEED EFFECTS ........................................................................... 59
Objectives ........................................................................................................................... 60
Field Testing ....................................................................................................................... 60
Test Section ..................................................................................................................... 60
Testing Dates and Speeds ............................................................................................... 60
Field Results and Discussion .............................................................................................. 60
Strain Definition.............................................................................................................. 60
Strain vs. Speed ............................................................................................................... 61
Strain vs. Temperature .................................................................................................... 62
Combined Effect .............................................................................................................. 62
Application of Regression Models .................................................................................. 63
Summary ............................................................................................................................. 65
FIELD-BASED PERPETUAL PAVEMENT STRAIN THRESHOLDS .............................. 66
Objective and Scope ........................................................................................................... 67
2000 Test Track .................................................................................................................. 67
2003 Test Track .................................................................................................................. 72
2006 Test Track .................................................................................................................. 74
Threshold Development ...................................................................................................... 79
Conclusions ......................................................................................................................... 80
CHAPTER 6 - LABORATORY AND FIELD COMPARISONS ......................................... 82
GRANULAR MATERIALS LABORATORY AND FIELD COMPARISONS ................... 82
Unbound Materials Used in the Structural Study ............................................................... 82
Backcalculation Cross-Section Investigation ..................................................................... 87
Field Characterization of Unbound Materials..................................................................... 90
Methodology ................................................................................................................... 90
Model Generation ........................................................................................................... 91
iv
Model Calibration........................................................................................................... 93
Summary of Field-Calibrated Stress-Sensitivity Models ................................................ 96
Comparison of Laboratory and Field Results ..................................................................... 97
Summary of Findings from Laboratory and Field Modulus Comparisons .................... 99
Recommendations ............................................................................................................. 100
E* LABORATORY AND MODEL COMPARISONS ....................................................... 101
Objectives and Scope ........................................................................................................ 101
Models to Determine E* ................................................................................................... 101
Witczak 1-37A E* Predictive Equation ............................................................................ 102
Witczak 1-40D E* Predictive Equation ............................................................................ 103
Hirsch E* Predictive Model .............................................................................................. 104
Testing Protocol ................................................................................................................ 105
Mixtures Tested ................................................................................................................ 106
Results and Discussion ..................................................................................................... 107
Summary ........................................................................................................................... 110
LABORATORY FATIGUE THRESHOLDS AND FIELD-MEASURED STRAINS ....... 110
Laboratory Testing ............................................................................................................ 111
Laboratory Testing and Field Data Comparisons ............................................................. 111
Phase One Comparison of Fatigue Threshold to Modeled Strain Levels ................. 112
Phase One - Analysis Results........................................................................................ 113
Phase Two Location of Fatigue Thresholds on Cumulative Distribution Plots ........ 114
Phase Two Analysis Results ....................................................................................... 115
Phase Three Ratio Comparison ................................................................................. 117
Phase Three Analysis Results .................................................................................... 118
Fatigue Threshold Summary ............................................................................................. 119
CHAPTER 7 FINDINGS AND STATE DOT IMPLEMENTATIONS ........................... 120
Mechanistic-Empirical Pavement Design ......................................................................... 120
Alabama ............................................................................................................................ 120
Florida ............................................................................................................................... 120
Georgia .............................................................................................................................. 121
Indiana............................................................................................................................... 122
Mississippi ........................................................................................................................ 122
Missouri ............................................................................................................................ 124
North Carolina .................................................................................................................. 125
Oklahoma .......................................................................................................................... 125
South Carolina .................................................................................................................. 126
Tennessee .......................................................................................................................... 127
Texas ................................................................................................................................. 128
CHAPTER 8 2009 TEST TRACK PLANS....................................................................... 129
LIST OF ACRONYMS
AASHTO: American Association of State Highway and Transportation Officials
ALDOT: Alabama Department of Transportation
AMPT: Asphalt Mixture Performance Tester
APA: asphalt pavement analyzer
APT: accelerated pavement testing
ARAN: Automatic Road Analyzer
ASTM: American Society for Testing and Materials
CAM: crack attenuating mix
CDF: cumulative distribution function
CTM: circular texture meter
DFT: dynamic friction tester
DGA: dense-graded asphalt
DSR: dynamic shear rheometer
E*: dynamic modulus
ER: energy ratio
ESAL: equivalent single axle load
ETG: expert task group
FDOT: Florida Department of Transportation
FHWA: Federal Highway Administration
Fn: flow number
FWD: falling weight deflectometer
GDOT: Georgia Department of Transportation
HMA: hot mix asphalt
HWTD: Hamburg wheel-tracking device
HVS: heavy vehicle simulator
IDT: indirect tension
IFI: international friction index
INDOT: Indiana Department of Transportation
IRI: international roughness index
LA: Los Angeles
LWT: loaded wheel tester
M-E: mechanistic-empirical
MDL: maximum density line
MDOT: Mississippi Department of Transportation
MEPDG: mechanistic-empirical pavement design guide
MODOT: Missouri Department of Transportation
MTD: material transfer device
NAPA-REF: National Asphalt Pavement Association Research and Education Foundation
NCAT: National Center for Asphalt Technology
NCDOT: North Carolina Department of Transportation
NCHRP: National Cooperative Highway Research Program
NMAS: nominal maximum aggregate size
OBSI: on-board sound intensity
ODOT: Oklahoma Department of Transportation
OGFC: open-graded friction course
PEM: porous European Mixture
PATB: permeable asphalt treated base
QC: quality control
RAP: reclaimed asphalt pavement
RBL: rich bottom layer
RMS: root-mean square
RTFO: rolling thin-film oven
SBS: styrene-butadiene-styrene
SCDOT: South Carolina Department of Transportation
SGC: Superpave gyratory compactor
SMA: stone matrix asphalt
SPL: sound pressure level
SPT: Simple Performance Tester
SRTT: standard reference test tire
TDOT: Tennessee Department of Transportation
TXDOT: Texas Department of Transportation
TWPD: three-wheeled polishing device
ULIP: ultra-light inertial profiler
VFA: voids filled with asphalt
VMA: voids in mineral aggregate
WMA: warm mix asphalt
EXECUTIVE SUMMARY
The original National Center for Asphalt Technology (NCAT) Pavement Test Track was
built in 2000 in Opelika, Alabama where it has served as a state-of-the-art, full-scale, closedloop accelerated loading facility. The construction, operation, and research at the Test Track
are funded through a cooperative effort of agency and industry sponsors with individual
pavement research objectives. The Test Track was designed to test 46 pavement sections
(200 feet in length) with 10 million equivalent single axle loads (ESALs) of traffic over two
years. The test sections allow pavement engineers and researchers to study pavement
responses and distresses to make pavement design more economical and efficient. The Test
Track underwent its first reconstruction phase in the summer of 2003.
After the second phase of testing (i.e., the second 10 million ESALs) had been completed,
twenty-two test sections at the Test Track were either milled or removed down to the
subgrade to prepare for the third phase of testing at the Test Track. Reconstruction for Phase
III occurred in the summer and fall of 2006. Trafficking of the pavement began on
November 10, 2006 and ran until December 4, 2008. At this point, the Phase III test sections
had received 10 million ESALs of traffic, whereas the remaining Phase II test sections had
accumulated 20 million ESALs, and eight test sections remaining from the original
construction had accumulated 30 million ESALs.
During the experimental phase of the 2006 Test Track, two distinct large-scale studies were
performed that encompassed many smaller sectional analyses: the mixture performance study
and the structural study. The mix performance study encompassed sections that were
originally built in 2000 and left in place through three cycles of traffic to quantify mixture
rutting potential in addition to newly placed mill and inlay sections containing high
percentages of reclaimed asphalt pavement (RAP).
A RAP experiment was conducted using mill and inlay sections at the 2006 Track. This
study was designed to evaluate the practicality of producing high percentage (25+ %) RAP
surface mixtures. Six sections containing RAP at either 20 or 45% were compared to a
control section of virgin mix. The sections were found to perform favorably in both the
laboratory and field in terms of rutting and cracking.
The Florida Department of Transportation (FDOT) has been developing a methodology for
predicting top-down cracking potential in a mixture based on an energy ratio (ER). To
validate this methodology, two sections were sponsored by FDOT with varying energy ratios.
The section with a lower energy ratio (N1) was expected to crack first. While the section
with a lower energy ratio did crack first, a forensic investigation was completed to ensure
other potential sources were not the cause of this early cracking. The forensic analysis
validated the difference in cracking between the two sections was not caused by differences
in mechanistic properties or bond strength.
The Georgia Department of Transportation (GDOT) funded construction and testing in
sections N11, N12 and N13 to compare the construction and performance of permeable
surface mixes containing two different aggregate sources that were placed with conventional
and dual layer paving equipment. While all the sections were constructed well, the section
placed by the dual layer paver seemed to be the most effective in terms of drainage and noise
reduction.
One of the objectives of the NCAT Pavement Test Track study was to evaluate correlations
between laboratory permanent deformation measurements using the Asphalt Pavement
Analyzer (APA) and Flow Number (Fn) tests and field rutting performance for a variety of
mixtures. APA and Fn tests were conducted on the mixes placed at the Test Track, and they
were correlated to rut measurements taken from the field. Based on these correlations, rut
depth criteria were determined and proposed for future implementation.
The structural study actually began during Phase II (2003-2005), part of which was continued
in the 2006 Test Track. This specific study was designed to aid in the calibration and
verification of mechanistic-empirical (M-E) design concepts by embedding instrumentation
in the pavement structure to measure mechanistic responses in the pavement under dynamic
loading. The 2006 Test Track had eleven test sections as part of this study. Four sections
(N3, N4, N6, and N7) were left in-place from Phase II while six test sections (N1, N2, N8,
N9, N10, and S11) were reconstructed from the subgrade up. Section N5 received a mill and
inlay, to mitigate top-down cracking, so it could endure an extra cycle of trafficking.
Four studies in the structural section were developed to verify or challenge different M-E
design principles. The first study compared the load durations of strain measurements at the
base of the hot-mix asphalt (HMA) in a 14 inch pavement section to the predicted load
durations from the mechanistic-empirical pavement design guide (MEPDG). This study
found that the MEPDG over-predicted the load duration of the pavement by about 80%. This
could lead to a pavement structure being over-designed.
A second study examined a common laboratory concept, fatigue thresholds, and developed a
field-based perpetual pavement strain threshold. This concept was developed by creating
cumulative strain distributions for test sections at the Test Track. The test sections that were
part of the 2003 and 2006 structural studies were chosen to be a part of this study as well as
six sections from the 2000 Test Track which survived for at least 20 million ESALs. This
study found that the cumulative strain distributions of test sections that experienced fatigue
cracking were different from those sections that performed well. The findings support a
field-based strain criterion (a strain distribution) for flexible perpetual pavement design based
on data from sections that survived 20 million ESALs without fatigue cracking. These strain
distributions were also related to the laboratory measured fatigue endurance limit by the
development of a fatigue ratio.
The comparison of laboratory and field characterizations for granular materials was the focus
of a third M-E design study. Falling weight deflectometer (FWD) testing was conducted to
characterize the granular materials in the structural sections of the 2006 Test Track, and these
backcalculated moduli were compared with popular recommended models. The results of
this study showed that either the MEPDG model or the universal model provide the best fit to
laboratory resilient modulus data, and the universal model provided the best model fit to
backcalculated resilient moduli. Therefore, the multi-variable constitutive models are
recommended over the single-variable models.
A final M-E study compared laboratory measured dynamic modulus (E*) values to predicted
E* values from the three recommended models. At the current state of the MEPDG, Witczak
E* predictive equations, 1-40D and 1-37A, are employed to determine dynamic modulus
given volumetric, gradation and binder properties of a mixtures. In applying these models to
ten mixtures included in the 2006 Test Track structural study, neither model consistently
estimated laboratory dynamic moduli values to a high degree of accuracy. Because both the
1-37A and 1-40D models were found to be unreliable and the 1-40D model largely overpredicts dynamic modulus, it is recommended that the Hirsch E* model be used. It should be
used with caution however, as discrepancies at lower temperatures and/or higher frequencies
were found.
One of the greatest benefits the Test Track has to offer its sponsors is the ability to test new
pavement design concepts and technologies in a controlled environment. States such as
Florida, Alabama, Missouri, Mississippi and Georgia have used research at the Track to
develop or change construction practices, alter design specifications, or validate new crack
prediction concepts.
The 2009 NCAT Pavement Test Track plans to expand the current structural study by
developing a six section Group experiment. This experiment will be used to validate M-E
concepts as well as study the durability and mechanistic properties of Warm-Mix Asphalt
(WMA). A second RAP study will be conducted to investigate the affects including high
percentages of RAP in base mixtures. Trafficking should begin for Phase IV of the Test
Track in the summer of 2009 and conclude in the fall of 2011.
CHAPTER 1 - INTRODUCTION
As new asphalt technologies and design frameworks move from the laboratory and computer
modeling toward implementation, there is a need to validate the new materials, processes,
and design methods. While some organizations consider full-scale testing on actual in-service
roads, completing such research can be limited by the following factors: (1) field
performance evaluations often take many years (15-20) to complete, (2) it is often difficult
and unsafe to close lanes on in-service roads for inspection and testing, (3) Departments of
Transportation are reluctant to leave roads in service until failure occurs, (4) the public can
be intolerant to traffic delays due to road closures, and (5) changes in personnel and political
climates can compromise long-term experiments (1).
Because of these difficulties, accelerated pavement testing (APT) facilities have become
more popular. Today, several APT facilities are scattered across the globe developing and
testing cutting edge materials and paving techniques to build longer lasting and more cost
efficient pavements. One such APT facility, the NCAT Pavement Test Track located in
Opelika, Alabama, just finished its ninth year of HMA research under controlled, but live,
traffic.
The Test Track was constructed in 2000 as a 1.7 mile closed loop facility. The Test Track
was developed as a partnership between Auburn University (who purchased the land) and the
Alabama Department of Transportation (ALDOT) who built the original research
infrastructure. The cost of site development, construction of the Test Track pavement
foundation and buildings, and data infrastructure was funded by ALDOT. This made it
possible for other states to conduct cost-effective pavement research without startup cost.
The pooled fund research cooperative with 3-year sponsorship commitments that was chosen
as the original funding model for Track construction and operations is still in use today.
Construction of the inaugural Track (Figure 1.1) was completed in the summer of 2000 and
then subjected to 10 million ESALs of truck traffic through December of 2002. Built as a
perpetual pavement, the first cycle of testing was a study of surface mix performance for
forty-six 200 foot test sections (Figure 1.2). Many of the test sections were rebuilt in the
summer of 2003, this time with a combination of mill/inlay surface mixes and variable
thickness structural sections. 10 million ESALs were applied to the new sections and
sections which remained in place from the original track accumulated 20 million ESAL
loadings. Likewise, the 2006 Track was a combination of varied thickness structural sections
and mill/inlay surface mixes. The 2006 NCAT Pavement Test Track represents the third 3year research cycle.
FIGURE 1.1 The NCAT Test Track.
FIGURE 1.2 Layout of Test Track (1).
CHAPTER 2 - OVERVIEW OF THE 2006 TEST TRACK (PHASE III)
Upon the completion of the 2003 NCAT Pavement Test Track experiment, the sponsoring
agencies were consulted in preparation of the 2006 Test Track experiment. While the 2003
Test Track had twenty-two test sections (Figure 2.1) remaining in place from the original
2000 Test Track, only eight of original test sections (shown in white) were chosen to receive
a third cycle of 10 million ESALs. Sixteen sections built during the 2003 experiment (shown
in yellow) were left in-place, and twenty-two sections (in blue) were either reconstructed or
rehabilitated based upon the sponsors research goals.
FIGURE 2.1 2006 NCAT Test Track.

Six sections (N1, N2, N8, N9, N10, and S11) were fully reconstructed from the subgrade up.
Pavement materials were excavated to the desired depth from these sections so new
subgrade, base and HMA materials could be constructed to each sponsoring organizations
requested design thickness. These six sections, along with one mill and fill rehabilitation
(N5), comprised the new structural study for the 2006 Test Track. These seven sections,
along with four remaining from the 2003 experiment, brought the total of fully instrumented
pavement sections at the Test Track to eleven.
The remaining fifteen sections were milled and inlaid. These treatments ranged in milled
thickness from 1.25 to 4 inches. These sections were located in the outside (i.e. trafficked)
wheelpath. These fifteen sections were designed by state agencies to validate the effects of
air voids, RAP percentages, permeable surface treatments, and poor Los Angeles (LA)
abrasive aggregate on the field performance of the mix.
The 2006 Test Track was designed to perpetuate many of the same research objectives as the
2003 Test Track. Its main objective was to evaluate the field performance of experimental
pavement mixes and structures in the field. The structural test sections were designed to
validate and calibrate new transfer functions for M-E design, develop recommendations for
mechanistic-based material characterization, characterize pavement responses in rehabilitated
flexible pavement structures, and determine field-based fatigue thresholds for perpetual
pavements (2).
EXPERIMENTAL DESIGN
The pooled fund research cooperative that funded the 2006 Test Track was supported by a
total of 13 states, the Federal Highway Administration (FHWA), and Oldcastle Materials
Group. Experimental design, construction, trucking operations, laboratory performance
testing, field performance testing, and forensics were funded by sponsor fees. Completion of
all project tasks was the responsibility of NCAT under the direction of the sponsor oversight
group. Sponsors could either provide their own mix designs, or mix designs could be
developed by NCAT. Individual experiments were designed by each sponsor to best meet
their specific needs. NCAT also evaluated Test Track sections as a large pool of data in
order to draw conclusions that could be broadly applied to the industry.
ALDOT funded research in sections N3, N4, N6, N7, S11 and joined with other sponsors in
the six section RAP experiment. Traffic was extended from the 2003 research cycle on
sections N3, N4, N6 and N7 to encompass perpetual pavement design concepts. A new
structural buildup was constructed in section S11 as part of the broad focus on general M-E
analysis and design. All ALDOT structural sections are supported by the stiff, low plasticity
metamorphic quartzite soil indigenous to the Track under 6 inches of dense crushed granite
base.
FDOT funded construction and testing in sections N1 and N2. Both sections were
constructed with fully instrumented structural buildups that were intended to complement
general work in M-E analysis and design. However, the primary objective of the experiment
was to validate FDOTs energy ratio method for predicting surface cracking. This research is
presented in greater detail in Chapter 3.
GDOT funded construction and testing of sections N11, N12 and N13. The objective of the
GDOT study was to compare the construction and performance of permeable surface mixes
containing two different aggregate sources that were placed with conventional and dual layer
paving equipment. All three sections were placed on perpetual foundations to ensure that
distresses would be isolated to the experimental surface mixes. This research is also
presented in greater detail in Chapter 3.
The Indiana Department of Transportation (INDOT) funded construction and testing of
sections S7 and S8. The objective of the INDOT experiment was to quantify the relationship
between air voids measured in quality control (QC) volumetric samples and rutting
performance in mixes produced with unmodified binder. Specifically, it was hoped that
lower boundary (i.e. air void percentage where removal and replacement would be
mandatory) could be identified. Lower levels of air voids were achieved by increasing the
asphalt content and adjusting the gradation in accordance with guidance provided by INDOT.
Because the mix produced in this study contained the same materials mixed in different
proportions, it was decided to double the scope of the experiment by splitting sections S7 and
S8 in half.
The Mississippi Department of Transportation (MDOT) funded construction and testing of
sections S2 and S3. The original section S2 surface mix (the top 1 inches) from the 2000
Track was milled and stored for use as RAP in the new section S2. The 3/8 inch nominal
maximum aggregate size (NMAS) replacement mix was finer than the original inch
NMAS 2000 mix (33 percent passing the #16 sieve versus 23), with a lower level of design
gyrations (85 versus 100) and higher asphalt content (7.0 versus 4.9). Further, the 2000 mix
contained all virgin materials with 8 percent limestone, while the 2006 mix contained 15
percent RAP and no limestone. Both the old and new mixes were produced using styrene
butadiene styrene (SBS)-modified PG 76-22. Reconstruction in section S3 consisted of the
placement of 1 inches of 75 gyration 3/8 inch NMAS stone matrix asphalt (SMA) mix
containing 70 percent limestone under 1 inch of 50 gyration permeable surface mix
containing 100 percent gravel. Mix in both S3 lifts was produced with SBS-modified PG 7622.
The Missouri Department of Transportation (MODOT) funded construction and testing of a
fully instrumented M-E structural section in N10. The buildup consisted of 8 inches of HMA
over 4 inches of dense crushed stone base over stiff Track subgrade. Although no direct
experimental control is available for this section, the similar buildup of Alabamas S11 does
provide for comparability of N10 to the broader M-E experiment of which S11 is a part.
The North Carolina Department of Transportation (NCDOT) funded continued trafficking on
sections S9 and S10 and also contributed funding for the RAP experiment. Sections S9 and
S10 are two of the original sections from the 2000 construction. S9 is a fine-graded
Superpave mix with PG 67-22 binder; S10 is a comparison coarse-graded section.
The Oklahoma Department of Transportation (ODOT) funded construction and testing in
sections N8 and N9. Both sections were fully instrumented for M-E analysis; however, the
HMA thicknesses were selected to investigate perpetual pavement design.
The South Carolina Department of Transportation (SCDOT) funded traffic continuation in
section S1. Originally built for the 2003 research cycle on a perpetual foundation, the 1-3/4
inch thick surface in section S1 was built with a inch NMAS SMA using an aggregate
blend with a history of poor LA Abrasion values.
The Tennessee Department of Transportation (TDOT) funded research in sections S4, S5, S6
and E1 on the 2006 Track. Traffic was extended on sections S4, S5 and E1 from 2003.
Sections S4 and E1 represented TDOTs very first permeable surface mix and SMA design
efforts, respectively. Continuing traffic on the permeable surface for the 2006 research cycle
extended the scope of the study to encompass long term drainability performance while the
focus of traffic continuation on the SMA surface was durability.
The Texas Department of Transportation (TXDOT) funded construction and testing in
section S12. The objective of the TXDOT experiment was to evaluate the effectiveness of a
crack attenuating mix (CAM) at preventing reflective cracking on slab concrete pavement.
10
Construction of section S12 presented a challenge, since no concrete pavements have been
placed at the Test Track. To meet TXDOTs research needs, the surface of the original
section S12 was milled to a depth of 4 inches. This was necessary because the CAM mix
was to be overlaid with a conventional dense surface mix. After milling had been completed,
a large diameter masonry saw was used to cut the Tracks remaining perpetual foundation in
both the longitudinal and transverse direction, effectively creating 15 foot (in the longitudinal
direction) by 12 foot (full lane width) free standing slabs. Approximately half of the 200 foot
section was cut in this manner, with the other half uncut to serve as a control. Fine sand was
carefully poured into the saw cuts to prevent the cracks from healing, then the mixes
selected by TXDOT were placed.
FHWA funded construction and testing in sections N5 and contributed to the RAP
experiment. The foundation for section N5 was the 5 inch thickness of structural section
remaining after the top 2 inches of 2003s section N5 was milled for rehabilitation.
Oldcastle Materials Group was the only private sector partner to fully support a test section
in the 2006 research cycle. They contributed funding for the construction and testing of the
RAP experiment.
A summary of research within all 46 experimental sections on the 2006 NCAT Pavement
Test Track is provided in Table 2.1. A + was added by the binder grade to designate a mix
design with binder greater than optimum.
In addition to the primary sponsors, the Track also received supplemental financial support
from several different sources. Both Nebraska and Wisconsin provided funds that were used
to support general research activities. A portion of the financial endowment from the
National Asphalt Pavement Associations Research and Education Foundation (NAPA-REF)
that is used to partially fund NCATs mission was used to indirectly support project activities
at the Track. Additionally, Auburn University provided funding that was used to design and
deploy the high-speed wireless data acquisition network at the Track. While the data
infrastructure upgrade was an important step in protecting the safety of the NCAT research
team (enabling them to collect high-speed response data from the safety of the laboratory
rather than from the roadside), the enhanced wireless network also created multi-disciplinary
opportunities for other Auburn University researchers.
The cost for Track reconstruction and operations is minimized through the generous support
of several different equipment manufacturers who donated the use of equipment and
technical support. During construction, Astec Industries routinely provided equipment that
facilitates plant production as well as mix placement. Compaction and placement equipment
has been provided by Bomag Americas, Dynapac, Ingersoll Rand and Roadtec. Construction
materials were provided by Boral Material Technologies, the Blaine Companies, Dravo
Lime, Hanson Aggregates, Martin Marietta Aggregates, MeadWestvaco, Oldcastle Materials
Group and Vulcan Materials. Many other material supply companies donated materials
directly to state DOT sponsors. Equipment for mix and pavement quality testing has been
provided by CPN International, the Gilson Company, HMA Lab Supply, Instrotek, Transtech
Systems and Troxler Electronic Laboratories.
11
TABLE 2.1 2006 Research Cycle (Newly Reconstructed Sections Shown in Bold Type)
Sec
Num
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
Test Study
Surface Mix
Year of
Design
Specified
Total
Sec HMA (in)
Stockpile Materials
Completion Methodology
Binder
HMA (in)
E2
4
Calcined Bauxite
2003
Superpave
Epoxy
24
E3
4
Calcined Bauxite
2003
Superpave
Epoxy
24
E4
4
Granite
2000
Superpave
PG76-22
24
E5
2
Grn/Lms/Snd (45% RAP)
2006
Superpave
PG67-22
24
E6
2
2006
Superpave
PG76-22
24
E7
2
2006
Superpave
PG76-22s
24
E8
4
Granite
2000
Superpave
PG67-22
24
E9
2
Granite/Limestone/Sand
2005
Superpave
PG67-22
24
E10
2
2005
Superpave
PG76-22
24
N1
7
Granite/Limestone
2006
Superpave
PG67-22
7
N2
7
Granite/Limestone
2006
Superpave
PG76-22
7
N3
9
2003
Superpave
PG67-22
9
N4
9
2003
Superpave
PG76-22
9
N5
7
2006
Superpave
PG67-22
7
N6
7
2003
Superpave
PG67-22
7
N7
7
Georgia Granite
2003
SMA
PG76-22
7
N8
10
Oklahoma Granite
2006
SMA
PG76-28
10
N9
14
Oklahoma Granite
2006
SMA
PG76-28
14
N10
8
St Louis/Porphyry
2006
Superpave
PG70-22
8
N11
2.75
Georgia Granite
2006
OGFC
PG76-22
24
N12
2.75
Georgia Granite
2006
OGFC
PG76-22
24
N13
4
Georgia Granite
2006
OGFC
PG76-22
24
W1
4
Georgia Granite
2000
SMA
PG76-22
24
W2
4
Porphyry/Limestone
2000
SMA
PG70-22
24
W3
2
2006
Superpave
PG76-22
24
W4
2
2006
Superpave
PG67-22
24
W5
2
2006
Superpave
RA500
24
W6
1
Limestone/Gravel/Sand
2003
Superpave
PG76-22
24
W7
4
Nova Scotia Granite
2002
NovaChip
PG76-22
24
W8
1
North Carolina Granite
2003
NovaChip
PG70-28
24
W9
1
2003
Superpave
PG76-22
24
W10
4
Gravel/Limestone
2000
Superpave
PG76-22
24
S1
4
South Carolina Granite
2003
SMA
PG76-22
24
S2
1.5
Gravel/Sand (15% RAP)
2006
Superpave
PG76-22
24
S3
2.5
Gravel
2006
OGFC
PG76-22
24
S4
4
Limestone
2003
OGFC
PG76-22
24
S5
1.5
Gravel/Limestone/Sand
2003
Superpave
PG76-22
24
S6
1.25
Grv/Lms/Snd (15% RAP)
2006
Superpave
PG76-22
24
S7A
2
Granite-/Limestone+/Sand2006
Superpave
PG64-22
24
S7B
2
2006
Superpave
PG64-22++
24
S8A
2
2006
Superpave
PG64-22+
24
S8B
2
2006
Superpave PG64-22+++
24
S9
3
2000
Superpave
PG67-22
24
S10
3
2000
Superpave
PG67-22
24
S11
7
2006
Superpave
PG76-22
7
S12
4
Arkansas Granite
2006
Superpave
PG76-22
24
S13
4
Oklahoma Granite
2000
Superpave
PG70-28
24
E1
4
Tennessee Limestone
2003
SMA
PG76-22
24
Base
SubResearch
Material Grade
Objective(s)
Granite
Stiff
HVS PG67 Validation w/ High Friction Surface
Granite
Stiff
HVS PG76 Validation w/ High Friction Surface
Granite
Stiff
Performance of Coarse Gradation
Granite
Stiff
RAP Mix Design/Construction/Performance
Granite
Stiff
Granite
Stiff RAP Mix Construction/Performance w/ Sasobit
Granite
Stiff
Performance of Fine Gradation
Granite
Stiff
Evotherm Warm Mix
Granite
Stiff
Evotherm Warm Mix w/ Latex
Limerock Stiff
Fracture Energy & M-E Design
Limerock Stiff
Fracture Energy & M-E Design
Granite
Stiff
M-E Design Validation/Calibration
Granite
Stiff
Granite
Stiff
Granite
Stiff
Granite
Stiff
Stiff Sub Soft
Perpetual Pavement & M-E Design
Stiff Sub Soft
Perpetual Pavement & M-E Design
Limestone Stiff
Granite
Stiff
Drainable Mix w/ Cubicle Aggregates
Granite
Stiff
Drainable Mix w/ F&E Aggregates
Granite
Stiff
Twin Layer Drainable Mix w/ F&E Aggs
Granite
Stiff
Columbus Granite SMA
Granite
Stiff
SMA Aggregate Quality
Granite
Stiff
Granite
Stiff
Granite
Stiff
Granite
Stiff
Low Volume Road Preservation
Granite
Stiff
Surface Friction Restoration
Granite
Stiff
Granite
Stiff
Granite
Stiff
Durability of Coarse Gravel Mix
Granite
Stiff
High LA Abrasion Loss SMA Aggregates
Granite
Stiff
Lower Gyration Mix Performance
Granite
Stiff
100% Gravel OGFC Performance
Granite
Stiff
100% Limestone OGFC Performance
Granite
Stiff
Granite
Stiff
Granite
Stiff
Effect of Low QC Air Voids on Rutting
Granite
Stiff
Granite
Stiff
Granite
Stiff
Granite
Stiff
Performance of Coarse Gradation
Granite
Stiff
Performance of Fine Gradation
Granite
Stiff
Granite
Stiff
Rich Bottom Layer On Slab Pavement
Granite
Stiff
Performance of Early Superpave Mix
Granite
Stiff
100% Limestone SMA Performance
CONSTRUCTION
Previous Test Track experiments were developed, let, and administered by the ALDOT.
However, the 2006 Test Track was let and administered under a contract through Auburn
University. A competitive bidding process was used to select the contractor that produced all
the mixes for the project. Although the contract contained a provision to allow any statecertified contractor to install a portable plant onsite, the work was ultimately awarded to a
local contractor with only a 10 minute haul distance who submitted the low bid.
Job mix formulas submitted by research sponsors typically require the use of aggregates that
must be hauled from remote locations. Since the NCAT trucking fleet was not being used in
the off-traffic reconstruction cycle, it was temporarily reassigned to the open road to longhaul the necessary stockpiles. Because all the materials specified in the GDOT job mix
formulas for sections N11, N12 and N13 could be picked up and delivered within a single
work day, they were hauled with NCAT research tractors driven by staff drivers pulling
leased dump trailers. Stockpiles that required overnight travel were outsourced to local
trucking companies. After the necessary aggregate materials were stockpiled on the
contractors yard, trial and placement mix production could begin.
Due to storage limitations on the contractors yard, it was necessary to stage select truckloads
of material at storage locations adjacent to the track before they were needed for mix
12
production. In these cases, the contractor was required to then short haul material from the
Track back to the plant in a manner that facilitated the production schedule. A loader with an
operator and a dump truck were assigned to the Track for this purpose. A detailed mapping
process was utilized at both the track and the asphalt plant to ensure that stockpile confusion
did not lead to bad mix production runs.
Prior to placement on the Track, trial mixes were initially produced so that plant
proportioning could be adjusted as necessary and research sponsors could witness the mix
being placed. Trucks containing trial mixes were hauled to the Track and sampled as if they
were production mixes to facilitate laboratory testing and evaluation. When all test data
related to the trial mix were compiled, they were presented to the research sponsors so
changes to the plant settings could be incorporated into the subsequent production of mix for
placement in test sections.
A special sequence of events was needed at the plant to produce uniform truckloads of mix
that exhibited predictable properties. After calibrating the aggregate flow rate for each
unique stockpile, sufficient quantities of uncoated aggregates (usually less than a truckload)
were initially run through the plant to ensure a uniform gradation. Liquid asphalt was then
turned on, with enough material bypassed from the drag chain to ensure that uncoated
particles were not placed in the storage silo. The bypass chute was then closed, and the
coated material was run into the storage silo until the computer indicated that approximately
one truckload had accumulated. This material was discharged into a waiting truck and
moved away for disposal on the contractors yard.
At this point in the process, it was assumed that steady state conditions were attained and that
subsequent material run into the silo would be uniform in terms of aggregate gradation or
asphalt content. As soon as a truckload of material had again accumulated in the storage silo,
it was discharged into a haul truck using standard best loading practices and sent to the Track
laboratory for sampling and testing. If the mix was intended for placement on the Track, the
process was repeated until a sufficient quantity of material was available to lay the required
mat. If the mix was run for trial purposes, or if it was the last truckload necessary to lay the
required mat, another truckload of coated material was run into the silo before the flow of
liquid asphalt was stopped and the plant was shut down. The cold feed bins were unloaded,
and the plant was readied for the next test mix (calibrating aggregate flow rates, resetting
plant proportioning, etc).
The contractor was responsible for hauling mixed material back to the Track for placement as
either trial mix or production mix. A sufficient work force and equipment resources were
staged at the Track to support both operations which were completed sometimes only
minutes apart. A sample stand located behind the NCAT laboratory adjacent to the Track
was used to obtain representative samples from haul trucks prior to placement of mix on the
Track. These mixes were split down to sample sizes that were suitable for standard
volumetric analysis, asphalt content testing and gradation analysis. Compiled data were
shared with sponsor representatives who were onsite to oversee production and placement
activities, and mat placements were not accepted until they met with the sponsors approval.
13
The compaction effort was achieved by at least three passes of a steel-wheeled roller. The
roller had the capability of vibrating during compaction; however, this technique was not
used on every test section. After the steel-wheeled roller was removed from the pavement
mat, the contractor continued compacting the mat with a rubber tire roller until the desired
density was achieved.
While many of the construction efforts were inlays into milled sections, the construction of
six new structural sections required excavations to the subgrade material for new subgrade,
base and HMA materials to be placed. These sections were constructed by creating a ramp at
both ends of the section for machinery entry and exit. The ramps allowed material to be
excavated to the needed depth below the surface of the pavement. Some excavations were
over five feet in depth. Once the excavations were complete, granular materials were placed
and compacted before the contractor placed the HMA.
TRAFFIC
Trafficking at the 2006 Test Track was conducted in a similar manner to previous Test Track
experiments. Four triple flat-bed trailer trucks and one triple box trailer loaded the pavement
from 5:00 AM until approximately 10:40 PM Tuesday through Saturday. Trafficking began
on November 10, 2006, and ended December 4, 2008, after approximately 10 million ESALs
had been applied to the pavement structures.
Table 2.2 provides the axle weights for each of the five trucks under normal loading
conditions. There were occasions when either due to a specialized study or due to
mechanical malfunction trailers were removed from their given tractor. This left the tractor
pulling either a single flat-bed trailer or a combination of double flat-beds.
TABLE 2.2 Truck Weights for 2006 Test Track
Truck ID
1
2
3
4
5
Average
COV, %
Steer, lb
Axle 1
10,150
11,000
10,550
10,500
11,200
10,680
3.9
Tandem, lb
Axle 2
Axle 3
19,200
18,550
20,950
20,400
20,550
21,050
21,050
20,700
19,850
20,750
20,320
20,290
3.9
4.9
Axle 4
21,650
20,950
21,000
21,100
20,350
20,760
2.2
Axle 5
20,300
21,200
21,150
21,050
20,100
20,760
2.5
Single, lb
Axle 6
21,850
21,000
21,150
21,050
21,500
21,310
1.7
Axle 7
20,100
20,900
21,350
20,900
19,500
20,550
3.6
Axle 8
19,966
20,900
20,850
21,050
20,300
20,613
2.2
PERFORMANCE MONITORING
Every Monday, trucking operations were suspended on the Track so that surface condition
studies could be conducted to document performance of all experimental sections. Field
performance evaluations focused on the middle 150 feet of each 200-foot test section to
eliminate the effects of transitions near section ends.
Rutting, texture, and roughness were measured by using a high speed Automated Road
Analyzer (ARAN) van. Cracking was determined by manually inspecting the surface of the
test sections. The manually detected cracks are often so fine (low severity) that they cannot
14
be detected with automated systems. Crack maps were generated from these examinations to
quantify the extent of cracking and to monitor the progression of cracking.
15
CHAPTER 3 - MIXTURE PERFORMANCE COMPARISONS
For one-to-one comparisons between mixes, pavement engineers need locations where mixes
can be tested side-by-side in the same conditions. While there are some limitations that
engineers must be aware of when comparing test sections at the NCAT Test Track, HMA
mixes placed during the same research cycle are generally ideal for making mixture
performance comparisons given they are in similar locations.
Past Test Track research compared performance characteristics of modified versus
unmodified asphalt binders, fine versus coarse graded mixtures, SMA versus Superpave, and
the effects of other aggregate properties. In 2006, this objective was expanded by
introducing sections that had low air voids, high percentages of RAP, and permeable surface
mixes. The main purpose of these sections was to investigate rutting resistance and/or early
field durability.
VALIDATION OF THE ENERGY RATIO CONCEPT
A distress plaguing the many highways is top-down cracking. FDOT has reported that over
90% of crack-related distresses in the state are top-down in nature (3). The exact mechanism
initiating top-down cracking is a complex interaction of load, thermal, and aging affects on
the roadway, and researchers have failed to find a single mix property to help designers
eradicate this problem (4).
Due to the absence of material properties that could discern between cracked and noncracked pavements, Roque et al. (4) developed the concept of a fracture energy ratio (ER)
which could possibly predict top-down cracking potential. It had been proposed (5; 6; 7; 8)
that an indirect tension test could be used to evaluate cracking potential by quantifying an
asphalt mixtures energy dissipation and energy threshold. When the amount of energy
dissipated by a pavement reached or surpassed the given threshold, cracking would occur.
Therefore, if one were to design a pavement with a higher energy ratio, the pavement should
be more resistant to top-down cracking (4).
FDOT sponsored two sections (Figure 3.1) at the Test Track during the summer of 2006 to
help validate the energy ratio design concept. The first section (N1) was built with a PG 6722 binder while the second section used a PG 76-22 binder. The PG 76-22 binder was
predicted to increase the ER of the mixture, given the same aggregate structure and structural
design (9).
These two sections were subjected to the accelerated trafficking patterns defined earlier.
Once trafficking began, performance differences were quickly noted between the two
sections as cracking began to propagate through N1. Later in the experiment, N2 began to
crack as well. However, while differences in the cracking pattern were observed, it was
important to clearly identify differences in ER as the primary cause cracking performance
differences.
During construction, it was noticed that a material transfer device (MTD) operated on one of
the paving lanes while laying the mix for section N1. The vehicle picked up tack on its tires
16
which might have caused bond failure to occur in this section. Previous research at the Test
Track has found relationships between poor bond and premature cracking (10; 11).
Therefore, to completely validate the energy ratio concept, a forensic investigation and
mechanistic analysis was initiated (9).
PG 67-22
PG 76-22
Test Section
N1
0.0
Lift 1
2.2
N2
2.0
Depth From Pavement Surface, in.
2.0
1.9
Lift 2
2.0
3.3
Lift 3
3.1
4.0
6.0
8.0
Limerock Base
10.0
Moisture Content
Unit Weight
12.0
11.9%
127.4 pcf
13.0%
129.6 pcf
10.0
10.0
14.0
16.0
18.0
Moisture Content
Unit Weight
9.8%
132.2 pcf
A-4 Subgrade
8.3%
132.6 pcf
FIGURE 3.1 FDOT Test Sections (9).

Laboratory Energy Ratio Determination
The tests needed to calculate an energy ratio are laboratory indirect tension, creep strain and
resilient modulus tests. These tests were performed on mix design samples. The energy
ratios from duplicate tests for section N2 (i.e. the polymer modified section) were about 2.4
times higher on average than the energy ratios for N1; therefore, it was expected that N2
would be more resistant to top-down cracking (9).
Observed and Mapped Cracks
Cracking was first observed in section N1 on April 9, 2007. Approximately 200,000 ESALs
later, cracking was observed in N2. The cracks began transversely and soon interconnected
throughout the wheelpaths. This crack progression was not consistent with other forms of
top-down cracking; however, coring did confirm the cracking was only within the upper two
HMA layers (9).
17
The cracks that progressed through N1 were, initially, not very wide; however, the cracking
spread quickly throughout N1, and by January 28, 2008, the entirety of the lane was cracked.
A shallow mill and inlay was used to repair the cracking in N1 using similar materials, and
the same type of cracking was soon observed in the section again (9).
In-Situ Pavement Response
As will be described extensively in Chapter 4, asphalt strain measurements are collected
weekly under full-scale loading from asphalt strain gauges embedded at the bottom of the
HMA layer. These measurements are then compiled into a database where a representative
strain response (95th percentile strain) was calculated for a given axle type each day of
testing. Having these recorded mechanistic responses allows comparisons to be made
between test sections. This data was used to validate that differences in pavement response
did not cause section N1 to crack first (9).
Comparing the strain and temperature measurements of the two sections in question for
single axles (Figure 3.2), similar strain magnitudes despite having slightly different as-built
properties were observed. The strains in N1 become more erratic once cracking has
propagated throughout the section. Once cracking has damaged the top layer of the
pavement, the effective pavement depth has decreased causing an increase in the strain
magnitude. Finally, the strains in N1 remained greater than those for N2 even after the mill
and inlay occurred (9).
Core Observations
Cores were taken from both test sections to conduct a visual investigation into crack type.
The cores validated the hypothesis of top-down cracking. The cores taken from N1 had
cracks ranging from hairline surface fractures to cracks that propagated through the upper
two inches of the pavement. However, below the top two inches, the cores were fully intact
(9).
Bond Strength Testing
Bond strength tests were conducted on the cores using a Marshall device and test procedures
described elsewhere (9). Table 3.1 shows the bond strength results. While one core from N1
showed atypically low bond strength, an ANOVA statistical test (=0.05) showed that the
two sections did have statistically similar means (F-statistic 0.376, F-critical = 7.7). Thus,
different bond strength was not a contributing factor to N1s early cracking relative to N2 (9).
18
Cracking throughout N1
(6/18/2007)
First Cracking in N1
(4/1/2007)
N1 milled and inlaid

(2/1/2008)
2000
120
100
1600
1400
80
1200
1000
60
800
40
600
400
20
200
4/24/2008
2/24/2008
12/26/2007
10/27/2007
8/28/2007
6/29/2007
4/30/2007
3/1/2007
12/31/2006
11/1/2006
Date
N1-Strain
N2-Strain
N1-Temp, F
N2-Temp, F
FIGURE 3.2 Measured Strain and Temperature versus Date (9).

TABLE 3.1 Bond Strength Test Results (9)
Section Core Shear Stress at Failure, psi
1
84
2
154
N1
3
197
4
157
1
149
N2
2
163
3
186
Energy Ratio Conclusions
This study investigated two sections designed to validate the ER concept. The following
conclusions can be drawn from this research.
Section N2 was designed with nearly double the ER as section N1, and it proved to be
more resistant to top-down cracking.
The consistency between the two sections in both bond strength and strain magnitude
provided evidence that bond strength was not the cause of one sections early failure.
Cores taken from the sections confirmed that the cracking was confined to the top of
the pavement structure (9).
19
Temperature, F
Longitudinal Microstrain Under Single Axle
1800
LABORATORY AND APT TESTING OF MODERATE AND HIGH RAP CONTENT
MIXES
The interest in increasing RAP contents in HMA is driven by escalations in asphalt binder
prices and the desire to conserve non-renewable resources. Many highway agencies
currently limit RAP contents to 15 percent or less. The FHWA RAP Expert Task Group
(ETG) identified two reasons higher RAP contents are not allowed are (1) the lack of
documentation of high RAP mixes field performance and (2) the lack of research on the
selection of appropriate virgin binder grades (12). High RAP mixes were defined by the
RAP ETG as mixes containing 25 percent or more RAP by weight of aggregate.
To study both the constructability and performance of moderate and high RAP content
mixes, six test sections (Table 3.2) were incorporated into a RAP experiment during the 2006
Test Track. The six test sections were built using a 50 mm mill and inlay with RAP mixtures
as noted in the table. Beneath the RAP inlay is a 560 mm HMA structure on top of an
aggregate base and Track subgrade (13). The virgin control section was the mill/inlay placed
on section N5.
TABLE 3.2 Summary of Test Sections and Binder Test Data (13)
Virgin Binder
Virgin Binder + RAP
%RAP
True
Predicted
Recovered
PG Grade
Section %RAP* Binder**
Grade
Grade
Grade
78.1 -23.8
W3
20%
18.2%
PG 76-22
80.1 -22.4
78.1 -30.3
68.4-31.2
W4
20%
17.6%
PG 67-22
72.0 -28.6
74.2 -29.7
54.7-32.8
W5
45%
42.7%
PG 52-28
69.4 -25.8
74.1 -30.2
68.4-31.2
E5
45%
41.0%
PG 67-22
76.9 -25.1
80.9 -26.2
78.1-23.8
E6
45%
41.9%
PG 76-22
82.7 -20.7
85.5 -25.7
PG 76-22 +1.5%
E7
45%
42.7%
83.2 -20.6
85.7 -18.8
86.3 -24.3
Sasobit
68.4-31.2
N5
0%
0%
PG 67-22
68.4 -31.2
71.1 -32.4
*by weight of aggregates **by weight of binder
Quality control data showed slight deviations in both air voids and voids filled with asphalt
(VFA) in the three test sections in the West curve. The air void contents were approximately
2%, and the VFA percentages were about 10% higher than the design range for heavily
trafficked pavements. These discrepancies were due to differences in the gradations of the
RAP stockpiles used for design and production (13).
During construction, the 20% RAP test sections were easily compacted under the first few
roller passes. Compactability of the 45% RAP test sections was influenced by binder grade.
The sections with the softest binder exhibited the least compaction resistance. The two
sections which required the most compactive effort were the 45% RAP sections which
contained the PG 76-22 binder and the PG 76-22 + Sasobit. The Sasobit was added to aid in
compaction, not for the reduction in production in temperature. However, the additive did
not appear to improve the compactability of the RAP mat.
20
Field Performance
Four primary areas of concern were investigated in terms of the RAP sections performance:
rutting, texture change, roughness, and cracking. Figure 3.3 shows the measured rut depths
of the seven test sections in the RAP experiment upon the completion of trafficking. Minor
rutting occurred in most of the test sections with the exception of the 20% RAP, PG 67-22
test section. 8.5 mm of rutting was measured in this section. However, these rutting
measurements were taken after two very warm summers. Since very little future rutting is
expected in these sections, with the exception of the 20% RAP, PG 67-22 test section, the
RAP mixes performed well despite the low air voids and high VFA percentages (13).
9
8
Rut Depth (mm)
7
6
5
4
3
2
1
0
0% RAP,
PG 67
20% RAP, 20% RAP, 45% RAP, 45% RAP, 45% RAP, 45% RAP,
PG 67
PG 76
PG 52
PG 67
PG 76
PG 76
(Sasobit)
FIGURE 3.3 Field Rut Depths after 10 Million ESALs for the RAP Experimental
Sections.
Raveling has been linked to macrotexture changes in previous Test Track experiments. The
ARAN van measured macrotexture using a high-frequency profile laser in the center of the
right wheelpath. While two of the 45% RAP sections (PG 76-22 + Sasobit, PG 67-22)
exhibited the greatest increase in texture, the textural increase was within the range typically
observed under trafficking at the Test Track; therefore, these sections are expected to resist
raveling (13).
Roughness was measured using inertial profiler technology described in AASHTO R 43-07.
Over the course of testing, the 45% RAP sections international roughness index (IRI) values
remained relatively constant. Two sections (control and 20% RAP, PG 67-22) in the RAP
experiment showed gradual increases in IRI; however, further investigations have removed
21
this fault from the RAP mixtures. Both sections have showed increased rutting which would,
in turn, increase the roughness number measured by the inertial profiler (13).
Two sections in the RAP study have experienced minor cracking; however, the cracks in
these sections have been linked to either reflective cracking (45% RAP, PG 76-22+Sasobit)
or segregation during construction (20% RAP, PG 76-22). More details regarding the crack
mapping of the sections in the RAP study are given by West et al. (13).
Laboratory Testing
Laboratory testing was also conducted as part of the research to test rutting potential and
durability using popular mixture testing techniques. Cores were taken from the RAP
sections, and plant produced mix samples were also collected during the construction of these
seven sections for laboratory testing.
Laboratory-created specimens were tested in the APA to quantify the rutting potential of the
mix. Generally, the test results from the APA (Figure 3.4) validated the field performance of
the test sections with the exception of the control section. A Fishers least significant
difference test showed that the control test section had statistically higher APA rut depths
than the RAP mixes. This statistical analysis also showed the 20% RAP, PG 67-22 mix had
greater rut depths than the other RAP mixes.
Reheated samples were used to conduct dynamic modulus testing using the Asphalt Mix
Performance Test (AMPT). The temperatures of the test were 4.4, 21.1, and 37.8 C while the
frequencies were 0.5, 1, 2, 5, 10, 20 and 25 Hz. The results of the mixes (Figure 3.5)
generally follow the expected ranking of stiffness. The lone exception to this occurred when
the 45% RAP mix with a PG 52-28 binder was ranked the stiffest at the low temperature
(13).
A third test, beam fatigue, was run on reheated plant mix using AASHTO T 321-07. The
tests were conducted on long-term aged specimens at 20 C and a constant strain of 500
microstrain. Failure was considered a 50% reduction in stiffness with original stiffness
determined at the 50th loading cycle. The results are given in Figure 3.6 (13).
22
16.0
Rut Depth @ 8,000 Cycles (mm)
14.0
12.0
10.0
8.0
6.0
4.0
2.0
0.0
20% RAP
PG 76-22
20% RAP
PG 67-22
45% RAP
PG 52-28
45% RAP
PG 67-22
45% RAP
PG 76-22
45% RAP Virgin PG

PG 76-22 +
67-22
Sasobit
FIGURE 3.4 Rut Depths from APA Testing (13).
|E*|, MPa
100000
10000
20% RAP 76-22
20% RAP 67-22
45%RAP 52-28
45% RAP 67-22
45% RAP 76-22
45% RAP 76-22+ Sasobit
Control
1000
100
-4.0
-2.0
0.0
2.0
4.0
Log Frequency, Hz
FIGURE 3.5 Master Curves of Dynamic Moduli for the RAP Experimental Sections
(13).
23
350
Fatigue Life (1000 cycles)
300
250
200
150
100
50
0
20% RAP
PG 76-22
20% RAP
PG 67-22
45% RAP
PG 52-28
45% RAP
PG 67-22
45% RAP 45% RAP

Virgin
PG 76-22 PG 76-22 + PG 67-22
Sasobit
FIGURE 3.6 Plot of Beam Fatigue Test Results for the RAP Experimental Sections (13).
An ANOVA statistical analysis indicated statistical differences existed between all the
datasets. Tukeys pair-wise comparison was further used to confirm that a statistical
difference was observed between the 20% and 45% RAP test sections. The 45% RAP mixes
has a significantly lower number of cycles to failure than the 20% sections. However, binder
grades did not significantly influence the beam fatigue results; therefore, it suggests that the
binder volume plays a larger role in fatigue life than binder stiffness (13).
Summary
Additional laboratory work is being conducted to further analyze the effects of virgin binder
grade on RAP mixes. When considering the field and laboratory test results from this study,
the following conclusions can be inferred.
Overall binder stiffness has an impact on the compactability of RAP mixes in the
field.
Despite low air voids and high VFA, the RAP mixes performed well in the field at the
NCAT Test Track in regards to rutting.
The minor cracking in test sections in the RAP experiment was not related to the
structural properties of the RAP mixture.
With the exception of the virgin test mix, rutting results from the APA matched field
rut measurements.
Master curves show that binder stiffness greatly influences mix stiffness. Softer
grades of binder decrease the mix stiffness with could decrease a pavements
durability.
Differences in beam fatigue results appear to be more affected by binder volume
content than by binder stiffness.
24
Based upon laboratory and field data collected at the NCAT Test Track, there does
not appear to be a strong case for supporting the use of softer binder grades in high
RAP mixes (13).
LOW NOISE PAVEMENT STRUCTURES
An evaluation of tire-pavement noise was conducted with different surface mixes. Test
sections consisting of coarse- and fine-graded mixes, SMA, and single and double layered
open graded friction course mixtures were paved on the inside lane of the north and south
tangents at the Test Track. NCAT Report 07-02 (14) reports on the design and testing of
these test sections.
Low Noise Pavement Design
Materials
Seven different pavement sections were constructed on the inside untrafficked lanes on the
south tangent of the NCAT test track. Asphalt mixtures used for the sections included three
SMAs with varying nominal maximum aggregate sizes, two dense-graded asphalt mixtures,
a micro-surfacing and an open-graded friction course (OGFC) (Table 3.3).
TABLE 3.3 Mix Designs for Noise Study (15)
Section
S2
S4
S5
S6
S7
S8
S9
Layer 1
OGFC 4.75 DGA 9.5 DGA <4.75 SMA 4.75 SMA 9.5 SMA Micro
(2 in)
Layer 2
Existing Track Milled Surface
The various mixtures were paved to a thickness of two inches over existing dense-graded
asphalt (DGA), which was milled. All mixtures were comprised of a PG 76-22 SBS polymer
modified asphalt binder and granite aggregate from Georgia (15).
Surface Macrotexture Measurements
The macrotexture of the surfacing layers were evaluated using the circular texture meter
(CTM) and the ultra-light inertial profiler (ULIP). These techniques are reported elsewhere
(16). Replicate CTM texture measurements were completed at five random locations on both
surfacing layers. Statistics of the macrotexture measurements using the CTM are shown in
Table 3.4. ULIP texture measurements are documented elsewhere (15)
25
TABLE 3.4 Surface Macrotexture (MPD, mm) of Noise Sections Using the CTM (15)
Section
S2
S4
S5
S6
S7
S8
S9
4.75
9.5
<4.75
4.75
9.5
Description
OGFC
Micro
DGA
DGA
SMA
SMA
SMA
Mean Profile Depth,
1.19
0.11
0.12
0.10
0.15
0.58
1.04
mm
Standard Deviation,
0.26
0.02
0.02
0.05
0.03
0.15
0.25
mm
Sound Testing
Sound measurements on the sections included sound pressure and on-board sound intensity
(OBSI) evaluations using the NCAT CPX trailer as detailed in NCAT Report 07-02 (14).
Repeated sound measurements (at least 3 runs) were done on each section. Testing was done
with the Michelin standard reference test tire (SRTT) at speeds of 45 and 60 mph.
Sound intensity testing with a NCAT triple trailer truck was also completed by attaching a
frame around the rear wheels on the rear trailer. Sound intensity microphones could be
mounted to record leading-edge and trailing-edge measurements. Details of the configuration
used are given in NCAT Report 07-02 (14). Sound absorption measurements with the 6 inch
diameter impedance tube were done in the field. Window putty was used to seal the base of
the tube that is placed directly onto the tested surface during measurements. This procedure
was evaluated in lieu of coring specimens and testing these in the laboratory.
Sound Pressure and Intensity Measurements
The sound pressure and intensity measurements on the low noise sections were completed on
March 3, 2007. The mean air temperature during noise measurements was 69 F. Results of
the sound pressure measurements on the low-noise sections measured using the CPX trailer
with SRTT tires are illustrated in Figure 3.7. For the sake of brevity, only the A-weighted
global sound pressure levels (SPL) calculated by logarithmic addition of the sound levels
between the third octave band frequencies of 316 and 3981 Hz are reported. The data for the
sound pressure measurements collected for the three runs at the front and rear microphones
are tabulated elsewhere (15).
26
FIGURE 3.7 Summary of Global Sound Pressure Levels at 45 and 60 mph (15).
Figure 3.8 summarizes the sound intensity measurements on the low noise sections done
using the CPX trailer at speeds of 45 mph and 60 mph and the NCAT semi trailer at a speed
of 45 mph. The data for the sound intensity measurements collected for the three runs at the
leading- and trailing-edge microphones are tabulated by Smit and Waller (15).
The noise spectra from the CPX sound pressure measurements and sound intensity
measurements at speeds of 45 and 60 mph with the SRTT tires have been documented
elsewhere by Smit and Waller (15).
FIGURE 3.8 Summary of Global Sound Intensity Levels (15).
27
Findings
Macrotexture
Although macrotexture significantly influenced the sound measured at the tire-pavement
interface, the relationship between macrotexture and noise is not well defined. This might be
because macrotexture is masked by other influence factors such as the porosity and stiffness
of the pavement mixture. It was reported that surfaces with the coarser Porous European
Mixture (PEM) were noisier than those with the finer OGFC (16).
It is proposed that the surface profile as defined using the ULIP be used to differentiate the
positive and negative texture on the road surface. This may be done, for example, by
applying a root-mean square (RMS) filter to the surface profile as shown in Figure 3.9. This
would allow a differentiation of positive and negative texture profiles.
FIGURE 3.9 RMS Filter Applied to ULIP Surface Profile (15).

Sound Pressure
Figure 3.7 shows the mean sound pressure levels for the different mixtures tested at 45 mph
and 60 mph. Contrary to expectations, the sound levels on the smoother surface mixtures are
higher than those on a coarser surface texture. An observation made when examining these
smooth surface mixtures in the field was that they would squeak or generate a high pitch
noise if the rubber soles of ones shoes were dragged across the surface. Although this is not
an objective test, it indicates possible noise effects under rolling tires. A possible
consequence of this phenomenon is the unusual peak sometimes seen in the data. It can be
seen that for the smoother surface mixtures on sections S4, S5, S6 and S7 that in addition to
the typical peak in the data observed at 1,000 Hz, an additional peak in the sound level data is
apparent at a third octave band frequency of 1,585 Hz and 1,995 Hz at 45mph and 60 mph,
respectively. The peak at the higher frequency could be indicative of a tonal noise generated
on these smoother surfaces. One possible explanation for this observation is that thin binder
films coating the surface aggregates were still present as these pavements were not subjected
to traffic before testing. The sound pressure spectra also indicate that the smoother surface
mixtures generate lower sound levels at the lower frequencies but higher noise levels at the
28
higher frequencies compared to the coarser surface mixtures. This partly explains why these
mixtures appear noisier as the human ear is more sensitive to higher frequency sound levels
and the A-weighting factors effectively filter out the lower frequency sound levels. The
better performing mixtures in terms of sound pressure level are the OGFC mix on section S2
and the micro-surfacing on section S9.
Sound Intensity
Figure 3.8 shows the mean global sound intensity levels measured using the CPX trailer at
speeds of 45 mph and 60 mph and those measured from an NCAT semi trailer at 45 mph.
The sound intensity measurements with the CPX trailer rank the surface mixtures in terms of
noise in the same order as the sound pressure measurements but are 1 to 2 dB(A) higher as
has been observed in previous studies (14). As with the sound pressure measurements, a
secondary peak in the sound intensity spectra is apparent at frequencies of 1,585 Hz and
1,995 Hz at test speeds of 45 mph and 60 mph respectively. Particularly high low frequency
(316 Hz) sound levels are apparent in the sound intensity spectra plots. This is probably
related to external noise influences resulting from possible vibration of the OBSI bracket.
Sound Absorption
The sound absorption plots (15) indicate negligible sound absorption for the dense-graded
and SMA mixtures, as expected. Sound absorption is apparent for the open-graded mixture
on section S2 and the microsurfacing mixture on section S9, albeit that the absorption
coefficients on the latter mixture vary considerably depending on where on the mat the
absorption testing was done. Low but discernable levels of absorption are apparent from two
tests on the 9.5 mm SMA mixture on section S8, probably related to surface scatter of the
sound waves from the coarser surface texture. This is emphasized to highlight the
corresponding trend observed in the sound level data. Surface mixtures with higher sound
absorption characteristics (such as open-graded mixtures) generally present lower CPX trailer
noise levels. This does not, however, explain the lower noise levels observed for the denser
mixtures under the NCAT trucks. This finding suggests that noise levels generated by heavier
vehicles or higher tire pressures are perhaps more influenced by the surface macrotexture
present at the tire pavement interface in contrast to lighter passenger vehicles, whose noise
levels are more influenced by the porosity of the surface mixture. The interaction between
surface macrotexture and porosity and the influence of these parameters on noise generated
at the tire-pavement interface is not clear. Intuitively one could speculate that vehicle tires at
higher pressures and under heavier loads would apply greater horizontal forces to positive
texture presented from coarser surface mixtures at the tire-pavement interface compared to
lighter vehicles with lower tire pressures traveling at the same speed. Hence the lower noise
levels of the heavier vehicles on the smoother surfaces. Further research is needed, however,
to better understand these influences.
Summary
The noise levels generated at the tire-pavement influence on roads is related to both
the macrotexture and porosity of the surface mixture. The degree to which these
factors influence the noise levels appears to be related to the weight of the vehicle on
the road and possibly tire pressure. For lighter passenger vehicles, the porosity of the
29
surface and the corresponding degree of noise attenuation as observed from noise
absorption testing appears to be the more dominant influencing factor. For heavier
vehicles (or higher tire pressures), the macrotexture of the surface and the positive
texture presented at the tire pavement interface has a greater influence.
To better understand the nature of the factors that influence road noise, it is
recommended that additional research be done to better define the interaction at the
tire-pavement interface.
Having defined the positive and negative texture present at the tire-pavement
interface, these profiles may be used to investigate these influences separately. The
positive texture affords a degree of stiffness representing the obstacle in the tires
path that would generate noise (possibly related to the forces applied at the surface)
and the negative texture represents porosity that would serve to attenuate noise
levels.
TEST TRACK FRICTION
The test sections at the NCAT Test Track have been measured for friction properties since
inception of the 2000 Test Track. Similar to rutting, each new surface mixture placed on the
track initiates a new set of measurements for friction. In the Test Track database, there are
six unique sets of friction data.
19 surfaces trafficked from 2000 to 2002
All of the sections have been measured for friction properties with standard skid trailers
operated by the Alabama DOT. In general, the frequency of friction measurement was
monthly, but a few periods extended further due to scheduling conflicts with the skid trailers.
In addition to the standard skid trailer friction measurements, the Dynamic Friction Tester
(DFT) and CTM were used on the Test Track sections. These devices are small portable
devices with a footprint of approximately 18 inches by 18 inches that measure the friction
resistance and macro-texture of a defined circular path. The DFT and CTM are American
Society for Testing and Materials (ASTM) standard tests. The amount of DFT and CTM
measurements on the Test Track sections is lower and less consistent than the skid trailer
measurements.
To compare skid trailer measurements and DFT measurements, the international roadway
safety community developed the universal International Friction Index (IFI). Through
correlation studies, the IFI was intended to compute comparable friction values from various
friction measurement devices. Research papers examining the application of IFI have had
mixed results. The friction study at NCAT will utilize the controlled traffic conditioning and
multiple HMA surfaces at the Test Track to further examine the use of IFI.
30
A global analysis of the friction performance of the sections at the Test Track has not been
undertaken until recent months. NCAT research in this area is still in its early stages. Most
of the emphasis is on developing a laboratory conditioning and testing protocol. With a
practical laboratory protocol, highway agencies will be able to evaluate HMA mixture
surface friction performance in a matter of weeks, instead of the three to five years needed to
gather data from full scale field test sections.
NCAT is currently working on a second phase study to calibrate and validate a laboratory
conditioning and testing program using the three-wheel polishing device (TWPD), DFT and
CTM. The TWPD was developed in an earlier NCAT study. A preliminary step of the
current research is to better define the friction performance of the Test Track sections and
select appropriate sections for the laboratory study. This effort is underway.
The initial focus of the Test Track friction analysis is to define each test section based on the
sections long-term friction performance. The boundaries for the study are the friction
measurements between the 2-million ESAL and 10-million ESAL traffic levels. Defining
friction relative to ESALs is comparable to using the real parameter (wheel passes). The
three parameters that will be determined from this initial analysis are:
o Long-term friction loss trend (linear best fit)
o Seasonal impact (polynomial best fit)
o Measurement variation (distribution about the seasonal best fit)
The friction data from initial construction to 2-million ESALs accounts for the dramatic
change in friction associated with friction increase as the asphalt binder film wears away and
the steep friction loss due to initial aggregate polishing. Most of the Test Track sections
display this friction pattern. After the initial aggregate polishing, the surface friction
performance stabilizes. The initial analysis of the study focuses only on the long-term
(stable) friction performance. A generic friction performance curve is given in Figure 3.10.
31
Friction value
Measurement variation
Seasonal impact
Long-term friction loss trend
2-million ESALs
Cumulative traffic
10-million ESALs
FIGURE 3.10 Generic HMA Surface Friction Performance Curve.

CONSTRUCTION AND PERFORMANCE OF PERMEABLE SURFACE MIXES
Sponsors typically fund research on two or more sections to compare the performance of
paving alternatives. In Phase III, GDOT sought to compare the construction and performance
of permeable surface mixes containing two different aggregate sources and to compare
permeable surface courses placed with conventional and dual layer paving equipment.
Permeable mixes with two different aggregate sources were studied to compare the effects of
aggregate shape (percentage of flat and elongated particles) on drainage characteristics and
performance. Currently, Georgia specifications require that aggregates used for porous
mixtures have LA Abrasion values of 50 or less. Georgias flat and elongated specification
limits 5:1 flat and elongated particles to 10 percent. A favorable comparison between the
single and dual layer pavements could result in improved performance with respect to
drainability, noise, etc. Constructability issues with the dual layer paver during construction
were also of interest to GDOT.
Experimental Design
All three sections were placed on perpetual foundations to ensure that distresses would be
isolated to the experimental surface mixes. Two porous pavement sections were placed with
a conventional single layer paver, one containing cubical aggregate (15% 3:1 flat and
elongated in section N11) and the other containing a higher percentage of flat and elongated
particles (29% 3:1 flat and elongated in section N12). The third section (N13) was built using
the same slightly flat and elongated aggregate; however, the surface consisted of a thin 9.5
mm NMAS porous mix over the same 12.5 mm NMAS porous mix. The purpose of the 9.5
32
mm NMAS upper mix was to prevent debris from potentially clogging the void structure of
the lower 12.5 mm NMAS mix. Figure 3.11 shows the aggregate gradation of the sections.
N13-1 and N13-2 in the chart refer to the mixture used in the top and bottom layers of section
N13, respectively. The DGA mix (S11) has been shown for comparison purposes.
100
90
80
Percent Passing
70
60
50
40
30
20
10
0
0.075 0.150 0.300 0.600
1.18
2.36
4.75
9.5
12.5
19.0
Sieve Size
N12
N13-1
N11
S11
N13-2
FIGURE 3.11 Aggregate Gradations of the Sections.

Drainable surface mixes on sections N11 and N12 were supported by 38 mm of conventional
GDOT dense-graded mix, while section N13 was supported by 54 mm of the same densegraded mix. The planned thickness of the 12.5 mm NMAS porous mix on all three sections
was 32 mm; however, section N13 also included 16 mm of 9.5 mm NMAS porous surface
mix. The job mix formulas are documented elsewhere (14).
Placement
Sections N11, N12 and N13 were milled beforehand to depths that would facilitate the
planned mat placements. A conventional paver was used to place all the lifts in sections N11
and N12, as well as the lower lift of dense-graded mix in section N13. A European dual layer
paver with a German crew was provided by Dynapac for the simultaneous placement of both
drainable mixes on the surface of section N13. The dual-layer paving machine (Dynapac F300 C/S) was manufactured in Germany, disassembled, shipped to the Track then
reassembled for use. Before being used to place mix on the track, the newly reassembled
machine was first used to simultaneously place two lifts of uncoated aggregate materials at
an off-track location. The dual layer paver was then moved to section N13 for placement of
the experimental mixes on the surface of the track. No tack coat was necessary between the
two porous mixes.
Dual layer paver technology has been used in Europe for approximately 10 years (18; 19).
The lower layer hopper was reported to have a 45 ton capacity, while the upper layer hopper
was reported to have a 25 ton capacity. One of the challenges in building the dual layer
section was filling of the pavers upper hopper. Because the Dynapac material transfer
device designed to feed the upper hopper of the dual layer paver was not available, a
33
temporary ramp had to be built on the track so that the Roadtec SB2500 material transfer
device in use at the Track could reach the upper hopper. Compactive effort was adjustable
on both the lower and upper screeds, with the amount of effort controlled by the crew. The
reported cost of the unit used at the Track was between $1.2 and $1.4 million (Nittinger,
Unpublished Data).
After work on all test sections was completed, the NCAT profiler was used to assess the
smoothness of the experimental mats. As seen in Figure 3.12, the smoothness of section N13
(placed with the dual layer paver) compared well with the smoothness of sections N11 and
N12 (placed with a conventional single layer paver). Additionally, mat densities were
determined by volumetric analysis of cut cores. Air voids in the mat (shown in Figure 3.13)
generally met GDOTs expectations for porous surface mixes.
Laboratory Performance
Cores were also obtained for permeability testing and interlayer bond strength measurements.
Bond strength measurements are presented in Figure 3.14 while laboratory permeabilities are
shown in Figure 3.15. It is seen in Figure 3.14 that bond strengths between the dual layer
pavements were comparable to bond strengths between tacked layers. Further, it is seen in
Figure 3.15 that permeabilities measured for the porous mixtures were very good.
Field Performance
After 10 million ESALs, very little rutting had occurred in any of the three test sections.
Changes in roughness and macrotexture resulting from the application of traffic are presented
in Figures 3.16 and 3.17, respectively. The lower macrotexture results for Section N13 are
due to the smaller NMAS gradation used as the surface of that section.
1.0
0.9
0.8
0.7
IRI (m/km)
0.6
0.5
0.4
0.3
0.2
0.1
0.0
N11
N12
N13
GDOT Permeable Test Section
FIGURE 3.12 Post-Construction Smoothness Comparison (14).
34
30
25
In-Place Air Voids
20
15
10
0
N11-1
N12-1
N13-2
N13-1
GDOT Permeable Sublot
FIGURE 3.13 Post-Construction Air Voids for all Porous Surface Lift Sublots (1 =
Upper Lift, 2 = Lower Lift) (14).
35
2.00
1.75
Bond Strength (MPa)
1.50
1.25
1.00
0.75
0.50
0.25
0.00
Permeable Layer Interface
N11
N12
N13 Lower
N13 Upper
FIGURE 3.14 Interlayer Bond Strength Measurements from Mat Cores (14).
200
175
Permeability (m/day)
150
125
100
75
50
25
0
Permeable Layer
N11
N12
N13 Lower
N13 Upper
FIGURE 3.15 Laboratory Permeability Measurements from Cut Cores (14).
36
2.00
International Roughness Index (m/km)
1.75
1.50
1.25
1.00
0.75
0.50
0.25
0.00
0
1,000,000
2,000,000
3,000,000
4,000,000
5,000,000
6,000,000
7,000,000
8,000,000
9,000,000
10,000,000
Equivalent Single Axle Loadings

N11
N12
N13
FIGURE 3.16 Roughness Measurements as a Function of Truck Traffic (14).

2.0
1.8
1.6
Mean Texture Depth (mm)
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
0
1,000,000
2,000,000
3,000,000
4,000,000
5,000,000
6,000,000
7,000,000
8,000,000
9,000,000
10,000,000

N11
N12
N13
FIGURE 3.17 Macrotexture Measurements as a Function of Truck Traffic (14).

Noise
Sound level measurements were performed approximately quarterly following construction,
beginning in December 2006. Table 3.5 includes the date of the measurements, noise level
37
recorded, and the amount of traffic (cumulative) that each section received prior to the
measurement. Data from the fine-graded HMA surface on section S11 is provided as a
control comparison. Figure 3.18 shows the sound pressure measurements through this
roughly two -year cycle.
Table 3.5 Summary of the Noise Measurements and Traffic
Sound Pressure Level (SPL), dB(A)
Traffic
Date
(ESAL)
N11
N12
N13
15 Dec-06
410.690
96.1
95.1
89.0
7 Mar-07
1,412,107
94.9
94.4
88.5
22 May-07
2,520,987
94.8
94.5
90.0
14 Dec-07
5,163,668
96.0
95.5
90.9
17 Mar-08
6,215,771
95.6
94.8
90.5
9 Jan-09
10,017,560
96.3
94.6
91.3
S11
94.4
92.4
93.2
94.4
94.8
95.5
100.0
98.0
96.0
SPL, dB(A)
94.0
92.0
90.0
88.0
86.0
84.0
82.0
N11
N12
N13
S11
80.0
Dec-06
Mar-07
May-07
Dec-07
Mar-08
Jan-09
FIGURE 3.18 SPL of the Sections over Time.

Section N13, the double-layer porous surface, shows a very gradual increase in sound
pressure level over the two year period, but remains the quietest of the four sections. Higher
noise levels were measured on the fine-graded Superpave surface mix on S11, and the 12.5
mm NMAS single-layer porous surface mixes in N11 and N12. These results show some
variations in the noise levels over time which may include aggregate reorientation, surface
wear and aging affects, as well as possible seasonal effects.
To better understand the noise absorbing characteristics of these test sections, Figures 3.19 to
3.24 show noise spectra measurements over this time period. Porous surface course sections,
N11, N12, and N13, revealed typical bell-shaped spectra. The double-layer surface on
38
section N13 shows lower amplitude values at each frequency band compared to the singlelayer porous surface sections. Section S11 was characterized by a spectrum shifted toward
higher frequencies which are more annoying to the human ear.
Analysis of the noise-frequency spectra for the double-layer porous surface, N13, through
each plot shows the low frequency (less than 1000 Hz) sound pressure measurements are
consistently lower than those of the N11 and N12 spectra.
Comparison of the noise spectra from sections N11 and N12 shows that the results were
similar except between 800 and 1600 Hz, where N12 (built with a higher percentage of flat
and elongated particles) was consistently quieter. This difference may be attributed to the
slight differences in particle shapes and gradations for these two porous friction courses.
However, the average difference over that frequency range was 0.8 dB (A) which is not
significant within human auditory perception.
Dec 06
95.0
N11
N12
90.0
N13
S11
SPL, dB(A)
85.0
80.0
75.0
70.0
65.0
60.0
315
400
500
630
800
1000
1250
1600
Frequency, Hz
FIGURE 3.19 Noise Spectra Dec-06 Measurements.
39
2000
2500
3150
4000
Mar 07
95.0
N11
N12
90.0
N13
S11
SPL, dB(A)
85.0
80.0
75.0
70.0
65.0
60.0
315
400
500
630
800
1000
1250
1600
2000
2500
3150
4000
Frequency, Hz
FIGURE 3.20 Noise Spectra Mar-07 Measurements.

May 07
95.0
N11
N12
90.0
N13
S11
SPL, dB(A)
85.0
80.0
75.0
70.0
65.0
60.0
315
400
500
630
800
1000
1250
1600
Frequency, Hz
FIGURE 3.21 Noise Spectra May-07 Measurements.
40
2000
2500
3150
4000
Dec 07
95.0
N11
N12
90.0
N13
S11
SPL, dB(A)
85.0
80.0
75.0
70.0
65.0
60.0
315
400
500
630
800
1000
1250
1600
2000
2500
3150
4000
Frequency, Hz
FIGURE 3.22 Noise Spectra Dec-07 Measurements.

Mar 08
95.0
N11
N12
90.0
N13
S11
SPL, dB(A)
85.0
80.0
75.0
70.0
65.0
60.0
315
400
500
630
800
1000
1250
1600
Frequency, Hz
FIGURE 3.23 Noise Spectra Mar-08 Measurements.
41
2000
2500
3150
4000
Jan 09
95.0
N11
N12
90.0
N13
S11
SPL, dB(A)
85.0
80.0
75.0
70.0
65.0
60.0
315
400
500
630
800
1000
1250
1600
2000
2500
3150
4000
Frequency, Hz
FIGURE 3.24 Noise Spectra Jan-09 Measurements.

Drainage
Observations of the drainage performance of the porous mixes were made at times when
natural rainfall presented the opportunity. Following extended rainfall events, videos were
made while following one the Tracks heavy trucks as it drove across all 46 test sections.
Still pictures clipped from a road spray video file recorded on May 15, 2008 after four hours
of rain falling at a rate of approximately 6 mm per hour are shown in Figure 3.25. Here, it is
seen that the rain has exceeded the drainage capacity of the porous mix on N11 containing
the most cubical aggregates. Water can be seen ponding in the wheelpaths. In comparison,
the porous mix on N12 containing a higher percentage of flat and elongated particles appears
to be nearing capacity, with isolated areas of visible water. The dual layer porous surface
appears to still be functioning very well.
FIGURE 3.25 Road Spray Observed Following Four Hours (6 mm per hour) of Rain
During Fleet Operations (14).
42
Conclusions and Recommendations
The following conclusions and recommendations can be made based on the results of this
comparison study that will be useful to both the highway industry and roadway construction
practitioners:
The construction quality of all three sections was generally very good, with mat air
voids and roughness values approximately equal. Besides the ramp improvisation
necessary for the MTD boom to load the upper hopper, no special problems were
encountered with the dual layer paver.
The interlayer bond strengths for all the porous mixes were excellent, including the
bond strength between the 9.5 and 12.5 mm dual layer lifts in section N13.
Laboratory permeabilities were also very good for all test mixes.
The field drainability exhibited by the section built with slightly flat and elongated
aggregate appeared to function better than the section built with the more cubical
aggregate after prolonged rain.
The field drainability exhibited by the dual layer pavement appeared to be better than
the conventional single layer drainable surface after prolonged rain.
At lower noise frequencies, the porous layer with more flat and elongated particles
showed lower sound amplitudes than the section with more cubical aggregate. This
aspect may be related to the volume and shape of the voids and their capabilities in
limiting resonance effects. These hypotheses require additional studies specifically on
the texture and its variability on the three porous surface sections.
Drainability, durability (as indicated by change in MTD) and sound should be
monitored with additional traffic in order to document any changes in performance
that could result from the accumulation of debris within the internal void structure of
these three porous pavement test sections.
DETERMINATION OF RUT DEPTH CRITERIA FOR LABORATORY
PERMANENT DEFORMATION EVALUATION
The Superpave mix design procedure was originally envisioned to consist of a volumetric
design procedure and performance tests (20). The volumetric mix design process has been
adopted by most state agencies, and the Superpave gyratory compactor (SGC) has been used
to determine volumetric properties for the general acceptance of HMA mixes. However, the
performance tests using the Superpave shear tester and the indirect tensile tester developed
for use in the Superpave mix design procedure are expensive and require extensive operator
training. Hence, these tests have not been adopted by state agencies for routine testing (21).
Increasing truck traffic has contributed to the demand for rut-resistant HMA on major
highways. Therefore, there is a need for a test that can indicate the rutting resistance of mixes
and that can be conducted in a timely manner during mix design processes or for quality
assurance purposes. For these reasons, loaded wheel testers (LWT), such as the APA and
Hamburg Wheel Tracking Device (HWTD), have been adopted for checking the rutting
resistance of HMA during the mix design process in many states.
The LWT rut depth has a reasonable correlation with field performance on an individual
project basis. However, prediction of field rut depths using LWT results for a given project
43
based on relationships developed from other projects with different geographic locations and
traffic is not generally reliable (22). In addition, LWTs do not provide a fundamental
property that can be used with an advanced material characterization model to account for
any specific stress state and temperature. Thus, research has been initiated to develop a new
performance-related test for permanent deformation for incorporation in the Superpave
volumetric mix design method and pavement structural design.
Under NCHRP Project 9-19 (23), the dynamic modulus (E*) and flow number (Fn) tests were
selected as simple performance tests for permanent deformation. However, the E* test has
been used mainly as the prime test method for HMA materials characterization for pavement
structural design in the Mechanistic-Empirical Pavement Design Guide (MEPDG) developed
under NCHRP Project 1-37A (24), and the Fn test has been considered as a potential
performance test for permanent deformation. Commercial testing equipment, the AMPT or
formerly known as Simple Performance Tester (SPT), is available for possible adoption of
the Fn and E* tests for routine use in the Superpave mix design method and in HMA
materials characterization for pavement structural design (25).
One objective of the NCAT Pavement Test Track study was to evaluate correlations between
laboratory permanent deformation measurements using APA and Fn tests and field rutting
performance for a variety of HMA mixtures used at the NCAT test track. Based on these
correlations, rut depth criteria were determined and proposed for future implementation.
Experimental Plan
The experimental plan is shown in Table 3.6. Ten test sections shown in the first column of
Table 3.4 were selected for this study. Sections S7 and S8 were divided into two 100-foot
subsections (S7A, S7B, S8A and S8B), and a unique HMA mixture was utilized for each
subsection. Thus, twelve plant-produced mixes were sampled.
Specimen Preparation
All mixes used for preparing specimens in the laboratory were plant-produced and sampled
during the construction of the test sections. Twelve mixes were sampled, of which nine
mixes were 12.5 mm nominal size and three mixes were 9.5 mm nominal size. Two test
sections used 15 percent RAP.
Table 3.7 shows the quality control gradation and binder content of the twelve mixes. A
detailed description of HMA mixture properties for each test section/subsection is available
on the NCAT Pavement Test Track website (26).
For each HMA mixture, twelve test specimens were prepared for the APA testing at the test
track laboratory immediately after sampling. Six replicates were compacted to Ndes gyrations
with a final height of 115 5 mm. These test specimens were not cut to a height of 75 1
mm but tested using molds that can accommodate 115 mm high compacted specimens. The
other six replicates were prepared for target air voids of 7 percent. The sample mass used for
preparing these specimens was calculated based on the research team experience with
mixtures used in the NCAT laboratory. Most of the samples had the air voids between 6 and
7 percent.
44
Three replicates were prepared of each HMA mix for Fn tests by reheating the plant-produced
mixes in the laboratory. For each test replicate, a cylinder 180 mm high and 150 mm in
diameter was first compacted in an SGC. The compacted specimen was then cored at the
center, and both ends of the core were trimmed. The final test specimen was 150 mm high
and 100 mm in diameter. The target air void level for these test specimens was 7 0.5%.
Table 3.8 shows a summary of the average and standard deviation of air voids for each set of
specimens tested in this study.
Laboratory Testing
All APA testing was conducted at 64 C (147 F). This is the high temperature of the standard
Superpave performance grade (PG) binder determined using the LTPP Binder software for
the test track location in Opelika, Alabama. After seating the test specimens, tests were run
for 8,000 cycles. A hose pressure of 830 kPa (120 psi) and a wheel load of 533 N (120 lb)
were used to test all of the specimens that were compacted to Ndes. These conditions were
recommended in NCHRP Report 508. For the group of specimens compacted to the target
air voids of 7 percent, a hose pressure of 690 kPa (100 psi) and a wheel load of 445 N (100
lb) were used in accordance with AASHTO TP 63-07. Both automatic and manual rut-depth
measurements were taken for each test.
The Fn tests were conducted using an AMPT. The specimens were tested at a deviator stress
of 70 psi and a confining pressure of 10 psi. The tests were terminated when the samples had
deformed to 100,000 microstrain. Some Fn tests were first conducted at the same temperature
as the APA test (64 C); however, the specimens failed quickly. Thus, the research team
decided to reduce the testing temperature from 64 C to 58 C. The flow number was
determined using two model formspower and Francken models. The non-linear regression
analysis used to fit both models to the test data was performed within the AMPT testing
software.
Test Results and Analysis
Table 3.9 shows (1) the average APA rut depths measured manually and automatically; (2)
the average flow numbers determined using the power and Francken models; and (3) the rut
depths measured in the field for approximately 5 and 10 million ESALs.
The correlations between the APA rut depths/flow numbers and field rutting performance
were shown in Tables 3.10 and 3.11 for 5 and 10 million ESALs, respectively. The most
common measure of correlation between data is Pearsons correlation (Pearsons r) with pvalue. Pearsons correlation reflects the degree of linear relationship between two variables.
It ranges from -1 to +1. A correlation of -1 or +1 means that there is a perfect negative or
positive linear relationship between variables. If the probability value (p-value) is lower than
5 percent (p-values < 0.05), the correlation coefficient (Pearsons r) is considered statistically
significant.
45
TABLE 3.6 Experimental Plan
Sect.
Sponsor
N1
N2
N8
N10
S2
S6
S7A
S7B
S8A
S8B
S11
S12
Mix
Type
Binder RAP% NMAS
FL
Super
67-22
FL
Super
76-22
OK
SMA
76-28
MO
Super
70-22
MS
Super
76-22
TN
411-D
64-22
IN
Super
64-22
IN
Super
64-22
IN
Super
64-22
IN
Super
64-22
AL
Super
76-22
TX
D-A
76-22
No. of Test Specimens
0
0
0
0
15
15
0
0
0
0
0
0
12.5
12.5
12.5
12.5
9.5
12.5
12.5
12.5
12.5
12.5
9.5
9.5
APA
Ndes
120lb /
120psi
6
6
6
6
6
6
6
6
6
6
6
6
72
7%
100lb /
100psi
6
6
6
6
6
6
6
6
6
6
6
6
72
Fn
7%
70psi /
10psi
3
3
3
3
3
3
3
3
3
3
3
3
36
TABLE 3.7 Quality Control Gradation and Binder Content of 12 HMA Mixtures
Sieve Size
1"
3/4"
1/2"
3/8"
No. 4
No. 8
No. 16
No. 30
No. 50
No. 100
No. 200
%AC
Percent Passing
N1 N2 N8 N10 S2 S6 S7A S7B S8A S8B S11
100 100 100 100 100 100 100 100 100 100 100
100 100 100 99 100 100 100 100 100 100 100
97 97
93
96 100 96
98 100 97
98 100
82 85
71
83
99 88
90
91
88
88 100
59 61
31
52
76 69
71
76
66
63
86
49 50
22
33
48 49
58
53
53
49
67
39 39
17
21
33 36
45
42
41
38
52
30 31
15
14
24 26
32
30
30
28
37
22 23
13
9
13 14
18
17
18
18
21
14 15
12
7
8
9
12
11
11
12
13
8.8 9.6 10.5 5.4 6.0 7.3 8.0 7.4 7.5 7.8 8.6
4.9 4.8 6.9 5.6 7.4 6.3 6.5 6.1 6.2 6.1 6.9
46
S12
100
100
100
97
65
38
36
18
11
10
7.4
7.4
TABLE 3.8 Average and Standard Deviation of Air Voids
Section
Ndes
N1
N2
N8
N10
S2
S6
S7A
S7B
S8A
S8B
S11
S12
Avg.
2.1
2.5
5.4
5.0
2.6
2.1
0.8
2.6
2.0
1.1
3.8
1.9
APA Test
7% Target Voids
Std. Dev.
0.08
0.14
0.38
0.28
0.15
0.27
0.10
0.26
0.13
0.10
0.10
0.20
Avg.
6.4
6.4
6.3
6.6
6.4
6.0
6.1
6.5
6.5
6.3
6.9
6.7
Std. Dev.
0.29
0.24
0.27
0.46
0.14
0.36
0.10
0.16
0.17
0.33
0.23
0.34
Fn Test
7% Target Voids
Avg.
7.0
6.9
6.7
7.0
6.9
6.7
7.0
7.0
7.0
6.8
6.8
7.0
Std. Dev.
0.30
0.30
0.30
0.15
0.40
0.29
0.17
0.35
0.31
0.10
0.06
0.20
As shown in Table 3.7, the APA automated measurement using 7 percent void specimens and
flow number determined using the Francken model exhibited reasonably good correlations
with the rut depths measured in the field for 5 million ESALs. These correlations were
significant (p-value < 0.05). The correlations between the laboratory and field rutting
performance for 10 million ESALs were not as strong as those for 5 million ESALs and not
significant (p-value > 0.05).
47
TABLE 3.9 Average APA Rut Depth Measurements and Flow Numbers
Section
APA Test
Ndes
120lb/120psi
Fn Test
7% Voids
100lb/100psi
7% Voids
70psi/10psi
Field Rut
~ 5 mil
~ 10
ESAL
mil
ESAL
Manual Auto Manual Auto Power Francken

(mm)
(mm)
(mm)
(mm) (cycle)
(cycle)
N1
3.76
2.55
3.65
4.99
493
493
1.40
N/A
N2
3.04
1.45
3.43
4.26
726
1121
1.50
3.7
N8
3.99
3.45
2.04
2.12
1877
1840
2.00
2.3
N10
4.74
2.99
3.85
4.42
500
618
8.30
12.3
S2
1.71
2.91
5.67
6.48
878
1110
0.70
0.7
S6
3.46
4.49
4.31
4.95
1615
1177
1.30
2.3
S7A
11.30
7.48
8.85
10.91
350
348
33.40
N/A
S7B
6.59
4.15
5.05
6.76
351
407
21.30
N/A
S8A
2.09
3.78
6.60
8.82
302
301
23.00
N/A
S8B
5.08
2.82
6.19
8.23
274
280
24.60
N/A
S11
7.26
5.71
5.93
6.89
656
752
6.80
11
S12
3.57
2.05
4.25
4.68
613
731
13.80
19.5
Notes: N/A = not available because the surface mixes of these sections were replaced after
approximately 5 million ESALs due to excessive cracking and rutting
TABLE 3.10 Correlations between Lab and Field Rutting Performance for 5 Million
ESALs
Correlation
Statistics
Pearson's r
p-value
APA Test
Ndes
120lb/120psi
Manual
0.612
0.035
Auto
0.499
0.099
Fn Test
7% Voids
100lb/100psi
Manual
0.751
0.005
Auto
0.795
0.002
7% Voids
70psi/10psi
Power
-0.637
0.026
Francken
-0.879
0.004
TABLE 3.11 Correlations between Lab and Field Rutting Performance for 10 Million
ESALs
Correlation
Statistics
Pearson's r
p-value
APA Test
Ndes
120lb/120psi
Manual
0.465
0.293
Auto
-0.08
0.865
Fn Test
7% Voids
100lb/100psi
Manual
-0.015
0.975
Auto
-0.060
0.898
7% Voids
70psi/10psi
Power
-0.623
0.135
Francken
-0.713
0.072
Figure 3.26 shows a comparison of rut depths measured in the field after 5 and 10 million
ESALs. For all of the sections that survived 10 million ESALs, the rate of rutting was higher
for the first 5 million ESALs from the start of traffic in November 2006 through the end of
2007.
48
Figure 3.26 also showed that if the critical rut depth in the field was 12.5 mm, all of the
mixes that performed well for the first 5 million ESALs would perform well in terms of
rutting for 10 million ESALs. In other words, the rut-susceptible HMA mixtures exhibited rut
depths that were greater than the critical rut depth of 12.5 mm within the first 5 million
ESALs. Thus, in order to determine critical rut depth criteria for rutting evaluation using the
APA and Fn tests in the laboratory, correlations between the laboratory and field rutting
performance for 5 million ESALs were developed. Figure 3.27 shows the correlation
between the average APA rut depths measured automatically using 7 percent void specimens
and the rut depths measured in the field. The correlation between the average flow numbers
determined using the Francken model and the field rut depths is presented in Figure 3.28.
40
Field Rut Depth (mm)
35
30
25
20
Critical Rut Depth in the Field = 12.5mm
15
10
5
0
N1
N2
N9
N10
S2
S6
S7A
S7B
S8A
S8B
S11
S12
Test Section
5 mil ESAL Rut
10 mil ESAL Rut
FIGURE 3.26 Comparison of Field Rut Depths after 5 and 10 million ESALs.
49
Conclusion
Based on the correlations shown in Figure 3.28 and 3.29, the following conclusions are
drawn:
The maximum acceptance criterion for APA tests should be 5.5 mm.
The maximum acceptance criterion for Fn testing should be 800 cycles.
These criteria were developed based on the laboratory test conditions presented in this
study and a conservative critical field rut depth of 3/8 inch (9.5 mm).
12.0
APA_7%_100lb/psi_Auto (mm)
10.0
APA Acceptance Criterion = 5.5 mm
8.0
r = 0.795, p-value = 0.002
6.0
4.0
2.0
0.0
0
9.510
15
20
25
30
35
40
Field Rut Depth (mm)

APA_7%_100lb/psi_Auto
Linear (APA_7%_100lb/psi_Auto)
FIGURE 3.27 Correlation between APA Rut Depths and Field Rutting Performance.
50
2000
Fn_7%_70psi/10psi (cycle)
1800
1600
1400
Fn Acceptance Criterion = 800 cycles
1200
1000
800
r = - 0.879, p-value = 0.004
600
400
200
0
0
9.5
10
15
20
25
30
35
40
Field Rut Depth (m m )

Fn_7%_70psi/10ps
Linear (Fn_7%_70psi/10ps)
FIGURE 3.28 Correlation between Flow Number and Field Rutting Performance.
LOW AIR VOIDS EXPERIMENT
The low QC air voids experiment sponsored by INDOT also yielded very interesting results.
INDOT reported in their research plan for the study that the implementation of volumetric
acceptance of HMA mixtures has exposed risks to the department and contractors. Prior to
this acceptance implementation, it was believed that mixtures placed on the roadway were
generally in accordance with design criteria; however, once volumetric acceptance was
implemented, problems involving the control of air voids began to surface.
Data from their QC program showed that the magnitude of variation for air voids was 4%,
four times greater than the 1% commonly believed. This means that in some cases
Superpave mixtures designed for 4.0 % air voids were placed with nearly 0 % air voids due
to construction variations. At the time the low QC air voids experiment was initiated at the
Test Track, mixtures placed with less than 2.0 % air voids risked removal and replacement
and mixtures placed with air voids between 4.0 % and 2.0 % were accepted with monetary
adjustments. INDOT was concerned with the performance of low air void mixtures left inplace while contractors were concerned with the monetary ramifications incurred through
removal and replacement.
The performance risk to INDOT versus the monetary risk to the contractors needed to be
evaluated. The risk of pavement failure for INDOT when HMA with less than 4.0 % air
voids is left in-place needed to be evaluated and compared to normal construction variation.
The monetary risk to contractors needed to be evaluated to determine if low air void HMA
could remain in-place and provide acceptable performance, thereby, reducing the removal
51
and replacement risks and resulting in lower HMA costs. For example, the following
outcomes were needed to address these issues:
Development of a rutting model for asphalt pavements with air voids between 0 %
and 4 %,
Validation of tests for assessing the rutting potential of HMA, and
Development of an acceptance pay structure for low air void mixtures based on
performance models.
To address this need, two 200 foot test sections were divided into four 100 foot subsections.
Transitional paving was accomplished by emptying the paver hopper at each marked 100
foot point and refilling with new mix containing slightly different aggregate and binder
blends. Three sections were produced with low QC voids by increasing the asphalt content,
and one section was produced by increasing the asphalt content and adjusting the aggregate
blend percentages. Another test section built with the same materials and similar gradation
was used as an experimental control. All mix was produced using an unmodified PG 64-22
asphalt binder (designated as a PG 67-22 in some states).
The first four subsections were placed in the fall of 2006 and performed well under traffic
throughout the winter and early spring. As seen in Figure 3.29, significant rutting was
observed when pavement temperatures increased in May of 2007 (after approximately 2.4
million ESALs). Rutting was allowed to progress throughout the summer and into winter of
2008. In February of 2008 (at approximately 5.6 million ESALs), all four experimental
subsections were replaced with new mixes at QC air voids levels that were intended to better
define the relationship with rutting performance. Rutting was again observed beginning in
May 2008 (at approximately 7 million ESALs).
Avg Rut Depth (mm)
11/9/2006
2/11/2007
5/17/2007
8/20/2007
11/23/2007
2/25/2008
5/30/2008
9/2/2008
12/6/2008
50
45
40
35
30
25
20
15
10
5
0
0
1,000,000
2,000,000
3,000,000
4,000,000
5,000,000
6,000,000
7,000,000
8,000,000
9,000,000 10,000,000

ARAN
Dipstick
ALDOT Gauge
FIGURE 3.29 Rutting Progression in Section S8B Showing Replacement Mix.

As seen in Figure 3.30, rutting for both the original and the replacement mixes progressed at
a more or less linear rate. In order to define the relationship between rutting performance
and QC air voids, the rate of rutting for the original and replacement subsections was plotted
against the corresponding QC air voids measured during construction. The plot also includes
52
results from the 2000 Track (also funded by INDOT) with slag and limestone mixes
produced with unmodified binder in which the asphalt contents were intentionally increased
by 0.5 percent. The resulting relationship is provided as Figure 3.30. Results from high RAP
content sections produced with unmodified virgin binder are also included in order to
demonstrate the broad relevance of these findings.
From a purely performance based perspective, it does not appear that any adjustment in pay
is required above approximately 2.75 percent QC air voids; however, the rutting rate
increases dramatically below this level and removal/replacement may be necessary. Please
note that Track data from the 2000, 2003 and 2006 research cycles all indicate that a
completely different relationship is expected for polymer-modified binders (with greater
tolerance for lower QC air voids).
4.0E-05
50
All mixes
mixes shown
shown produced
produced with
with PG67
PG67 binder
binder
-- All
Red sections
2000 (limestone/slag)
-- Green
control placed
sectioninplaced
in September 2006
Bluetreatment
sections placed
in placed
2006 (virgin)
-- Red
sections
in October 2006
Blackreplacement
sections placed
in 2006
(RAP)
-- Blue
sections
placed
in February 2008
- More rutting expected in blue replacement sections
- Design gradations shown with round symbols
- Closed gradations shown with triangle symbols
- Black diamonds show PG67 RAP study sections
- Yellow squares show 2000 limestone/slag mixes
Depth (mm)
RutESAL)
Measured
Maximum
/ Hot
(mm Avg
Rate ARAN
Rutting
45
3.5E-05
40
3.0E-05
35
2.5E-05
30
2.0E-05
25
20
1.5E-05
15
1.0E-05
10
-0.83x
y = 98.74e
2
2
R =R0.69
= 0.67
5.0E-06
0
0.0E+00
0.0
0.0
0.50.5
1.0
1.0
1.5
1.5
2.0 2.0
2.5 2.5
3.0
QC
AirVoids
Voids
(% Gmm)
QC Air
(% Gmm)
3.0
3.5
FIGURE 3.30 Relationship Between Rutting and QC Air Voids.
53
3.5
4.0
4.04.5
5.0
4.5
CHAPTER 4 - STRUCTURAL STUDY
The structural study at the NCAT Test Track began as part of the 2003 Test Track
experiment with eight instrumented test sections. These sections were designed to meet the
needs of M-E design validation by studying the interaction between pavement response and
performance for pavement structure constructed with different materials and thicknesses.
The study was expanded in the 2006 Test Track to eleven sections where four test sections
were left in-place from 2003, one section was rehabilitated, and six sections were completely
reconstructed. This chapter provides details about the pavement sections, instrumentation
used, mechanistic pavement properties, and pavement performance.
EXPERIMENTAL OBJECTIVES
To address the needs stipulated above, a number of broadly-defined experimental objectives
were developed for the 2006 Test Track structural experiment. These objectives, with brief
explanations, included (2):
Further validate and calibrate new transfer functions for M-E design.
On sections left in place from the 2003 experiment, initial transfer functions had been
developed. These models require further refinement and validation before they can be
widely applied. Additional calibration activities on new test sections will increase the
size of the calibration data set which can then be applied to a wider set of real world
design scenarios.
Develop recommendations for mechanistic-based material characterization that
yields accurate pavement response predictions.
To optimize pavement design it is important to accurately characterize the material
properties which directly affect predictions of pavement response under load. There
are currently many methods of material characterization in the context of M-E design.
These methods range from laboratory-based to field-based and include direct
measurement versus correlation equations. Furthermore, the methods can often
produce conflicting sets of information regarding the same material. Therefore, there
is a need to investigate the various methods and recommend best practices toward
mechanistic characterization of material properties.
Characterize pavement response in rehabilitated flexible pavement structures.
Many agencies are faced with rehabilitating flexible pavements, typically with
overlay or mill and inlay techniques. There is a need to validate pavement response
predictions made in rehabilitation design methodologies to improve and/or refine
these methods.
Determine field-based fatigue response thresholds for perpetual pavements.
Many state agencies have begun to design and construct so-called perpetual
pavements (a.k.a., long-life pavements). A critical component of the design process is
the selection of the fatigue threshold. Most laboratory fatigue testing of asphalt
mixtures have set a conservative strain threshold of 70 microstrain ( ) to prevent
54
bottom-up fatigue cracking. However, it is believed that the threshold may be much
higher due to rest periods and other differences between the field and lab. Pavement
response and performance measurements can help refine and update the strain
threshold for fatigue performance.
2006 TEST SECTIONS
As noted above, the sections comprising the 2006 structural study were a blend of new
construction, left-in-place and rehabilitated sections. Figure 4.1 illustrates the average asbuilt thicknesses and the constituent materials, in each section.
There are five different types of unbound materials that are utilized for the structural study at
the Test Track (27). A Florida limerock base was utilized as the base layer material in
sections N1 and N2. This material was quarried in Alachua, Florida, and is commonly
utilized by the Florida DOT. The granite graded aggregate base material supplied by Vulcan
Materials, Inc. was utilized as the base layer material in section N3, N4, N5, N6, N7, and
S11. This material is commonly used by ALDOT in the southeastern part of the state and was
quarried in Columbus, GA. The Type 5 material supplied by the Missouri DOT was used as
the base material in section N10. This material is a dolomitic limestone that was quarried in
Maryland Heights, Missouri, and is commonly used by the Missouri DOT. The Seale
subgrade material was employed by the Oklahoma DOT as the subgrade layer in sections N8
and N9. This material is high clay content borrow material imported from Seale, Alabama.
This soil is classified as an A-7-6 material by American Association of State Highway and
Transportation Officials (AASHTO) soil classification. Finally, the metamorphic quartzite
soil excavated from the Test Track property was utilized as the fill material in every section
except N8 and N9. This material was used as the base layer material for N8 and N9 to
simulate lime stabilization often used in Oklahoma and formed the deep subgrade material
for each structural section. This material is classified as an A-4(0) soil.
Florida DOT: Sections N1 and N2
FDOT sponsored two structural sections (N1 and N2) consisting of approximately seven
inches of HMA over ten inches of limerock base. The main difference between the sections is
that N2 included an SBS modified PG 76-22 binder in the upper four inches of HMA,
whereas N1 used an unmodified PG 67-22 throughout the full seven inches of HMA. The
upper lifts from the two test sections were designed to yield significantly different resistance
to cracking as indicated by their indirect tension (IDT) energy ratios. It was expected that
section N1 would crack before N2. Also, the introduction of a base material different from
the other structural test sections allows for interesting comparisons between the performance
of granular base layers.
Alabama DOT and FHWA: Sections N3 N7
As noted previously, sections N3 through N7 were left in place from the 2003 Test Track.
They all share the same high-quality subgrade and 6 inch unbound granite base layer.
Sections N3 and N4, both consisting of 9 inches of HMA, are companion sections with the
binder type as the main experimental variable. N5 and N6 were also companion sections
consisting of 7 inches of HMA, though N5 was milled 2 inches and inlaid with new HMA to
55
mitigate the top-down cracking problem. N7 was also left in place and features SMA as the
surface course.
Florida
(new)
N1
Oklahoma
(new)
Alabama & FHWA

(left in-place)
N2
N3
N4
N5
N6
N7
N8
FHWA
N9
Missouri Alabama
(new)
(new)
N10
S11
0.0
2.0
4.0
As Built Thickness, in.
6.0
8.0
10.0
12.0
14.0
16.0
18.0
20.0
22.0
24.0
PG 67-22
PG 76-22
PG 76-22 (SMA)
PG 76-28 (SMA)
PG 76-28
PG 64-22
PG 64-22 (2% Air Voids)
PG 70-22
Limerock Base
Granite Base
Type 5 Base
Track Soil
Soft Subgrade
FIGURE 4.1 As-Built Pavement Cross Sections (2).

Oklahoma DOT: Sections N8 and N9
ODOT also sponsored two new structural test sections (N8 and N9) intended to study the
perpetual pavement concept. These sections represent the thickest cross-sections built, to
date, as part of the structural experiment. The ODOT sections feature a soft subgrade which
was more representative of some soils in Oklahoma. Section N8, the first of the two ODOT
sections, has ten inches of asphalt which is made up of a two inch rich bottom layer, six
inches of Superpave mix, and capped with a two inch layer of SMA. The second section, N9,
has a total HMA thickness of fourteen inches. The rich bottom layer was increased to three
inches and an additional three inch Superpave lift was added for this section. It should be
noted that the so-called rich-bottom was simply a mixture designed to 2% air voids rather
than 4%. The net result was a 6% design asphalt content in the rich-bottom.
FHWA: Supplemental N9
Accounting for the viscoelastic nature of hot-mix is a critical component of any M-E design
methodology. Specifically, it is important to account for pavement temperature and loading
frequency (i.e., truck speed) when predicting pavement response using a mechanistic model.
While viscoelastic properties of individual mixtures can be readily measured in the
laboratory through dynamic modulus testing, translating the results into a form useful for
56
predicting pavement response can be complicated due to differences in mode of loading, rest
periods and other factors. There is an added level of complexity due to the layered nature of
flexible pavements with each layer potentially having different viscoelastic properties.
Furthermore, temperature and loading frequency change with depth through the layers so
establishing representative temperatures and frequencies can be particularly challenging.
Methods for handling the above complexities of mechanistic modeling have been welldocumented and even implemented within design programs such as the MEPDG.
However, there are a large number of simplifying assumptions that require validation to
ensure maximum accuracy within the design system. Therefore, the supplemental section
sponsored by FHWA was intended to provide a model validation data set for evaluating the
accuracy of various material characterization and modeling approaches currently available.
One feature of this investigation was embedment of strain gauges at lift interfaces to a
minimum depth of five inches to quantify pavement response with depth. N9 was selected
for this supplemental investigation because it was the thickest HMA section available and
allowed for the greatest strain and temperature profiles to be developed. The supplemental
instrumentation used in section N9 is shown in Figure 4.2. Full details regarding
instrumentation are provided later in this chapter.
Missouri: N10
The Missouri DOT has sponsored a structural section (N10) which was designed to address
the broad needs of mechanistic-empirical pavement design. N10 features a Missouri Type 5
granular base material beneath eight inches of HMA which are commonly used in Missouri.
Alabama DOT: S11
A new section was sponsored by ALDOT meant to build on previous work in the 2003
Test Track. In the 2003 structural sections, the same binder grade was used for all layers in a
given test section. In practice, however, agencies typically use higher binder grades near the
pavement surface where HMA rutting is more likely to occur and lower binder grades lower
in the pavement where temperatures are not as extreme and rutting less likely to occur.
Section S11 was built to more closely replicate typical pavement cross sections upper three
inches contained a PG 76-22 binder and the four inches contained a PG 67-22 binder.
57
0
1
SMA
Asphalt Strain Gauge
2
3
Thermistor
PG 76-28
Depth, in.
5
6
7
PG 64-22
Asphalt Strain Gauge
Plan View
PG 64-22
9
10
11
PG 64-22
Rich Bottom
12
13
14
CL
OWP
15
FIGURE 4.2 N9 Supplemental Instrumentation (2).

INSTRUMENTATION
The instrumentation system developed and deployed for the 2003 structural study had proven
itself to be reasonably robust and effective in gathering the requisite pavement response data
needed for M-E investigations. Therefore, essentially the same system, with a few
exceptions, was used for the 2006 investigation. This also maintained continuity between the
two research cycles in terms of equipment, data collection and data processing schemes.
This testing plan and methodology has been documented by Timm (2).
58
CHAPTER 5 - MECHANISTIC PAVEMENT ANALYSIS
A primary objective of the 2006 Structural Study was the analysis of mechanistic pavement
design concepts. This was accomplished by embedding instrumentation into the pavement
structure of eleven sections at the 2006 NCAT Test Track. Two major studies have been
completed and documented relating to different components of M-E design to date. The first
study synthesized in this chapter determined the effects that temperature and speed have on
the mechanistic responses of flexible perpetual pavements. The second study developed
cumulative distributions of strains for twenty-one test sections (six built in 2000, seven built
in 2003, and eight built in 2006) and compared the distributions to pavement performance.
Both studies have been documented in detail elsewhere (28; 29).
TEMPERATURE AND SPEED EFFECTS
M-E design procedures such as the MEPDG utilize material characteristics, climate, and
layer thickness to predict pavement responses under loading (30; 24). Transfer functions are
then employed to predict pavement performance based on the computed responses.
Responses that are essential to the successful prediction of load repetitions to failure include
the horizontal tensile strain at the bottom of the HMA layer to predict fatigue cracking, and
the vertical compressive strain at the top of the subgrade layer to predict rutting.
The horizontal tensile strain is of particular interest for the design of perpetual pavements
which have come to the forefront of the asphalt industry in recent years. One key feature of
these long lasting pavements is that each layer is designed to resist specific distresses. In
particular, the bottom layer is designed to resist bottom-up fatigue cracking by considering
the flexibility and thickness of the HMA (31). Fatigue cracking is dependent on the
horizontal tensile strain induced at the bottom of the HMA. If these strains are kept below a
maximum value it is believed that the pavement can sustain an infinite number of load
repetitions without failing by fatigue (31). Early estimates of the threshold were
approximately 70
while more recent observations have reported a threshold of 100
(32;
33). As a result of the zero damage incurred, the number of load repetitions that occur at or
below the strain threshold can be ignored in determining the load repetitions to fatigue
failure, ultimately reducing the required layer thickness. Therefore, properly predicting the
tensile strains at the bottom of the HMA is critical to designing a successful and costefficient perpetual pavement.
Recent research efforts at the NCAT Test Track have included investigations of perpetual
pavements. ODOT has sponsored perpetual pavement sections at the Test Track to examine
their behavior in an accelerated loading environment. One of the long-term goals of this
research is to characterize the performance of their perpetual sections so that the relevant mix
design and structural design procedures can be refined prior to widespread implementation
throughout Oklahoma. Central to this effort is characterizing a field-based threshold for
structural design. To accomplish this goal, both pavement response and performance require
characterization. As described below, this investigation focused primarily on pavement
response characterization.
59
Objectives
Given the needs regarding perpetual pavement design thresholds stipulated above, the
objectives of this study were:
Characterize the effects of temperature and speed on measured perpetual pavement
response.
Describe the strain history of the perpetual pavement in the context of existing
response thresholds.
Field Testing
Test Section
This experiment consisted of the application of live loads at a minimum of 4 different speeds
on 4 different dates to measure the strain induced throughout the N9 test section. During the
testing, temperatures throughout the pavement structure were also monitored. Section N9
was a perpetual pavement section and was comprised of approximately 14 inches of asphalt
concrete over a 9.6 inch granular base.
Testing Dates and Speeds
Testing was conducted on April 6, 10, 25 and May 2, 2007. On each date, the trucks traveled
at four speeds: 15 mph, 25 mph, 35 mph, and 45 mph. On the first two dates, data were also
gathered at 55 mph. However, for safety reasons, it was decided to discontinue testing at this
speed on the remaining dates (the Test Track is designed for 45 mph). For each speed tested,
the trucks adjusted to the test speed prior to traveling over the test section, such that once
over the test section, the speed was constant. It should be noted that section N9 was
subjected to routine traffic at 45 mph as part of the overall Test Track trucking operation.
The various test speeds used in this particular investigation were intended to provide a wider
range of conditions and therefore more applicability to potential lower-speed conditions of
actual roadways.
Field Results and Discussion
The results of the tests were processed and transformed into useable data to analyze the
pavements response due to the speed at which the load was applied and the in-situ
temperature of the pavement. The useable data was in the form of strain traces, which
plotted the strain () versus time. These strain traces allowed for the definition of the
induced strain under dynamic loading and for the analysis of strain due to vehicle speed and
pavement temperature.
Strain Definition
In investigating the effects of speed and temperature on strain within the pavement, focus
was placed on the tensile strain induced at the bottom of the HMA, where the maximum
tensile strain occurs. The tensile strains captured for all three passes of the trucks at a
particular speed and date were compiled separately by direction and axle type. Among the
strains captured from the six longitudinal and six transverse gauges, the maximum of the
strains induced under each axle type was selected for the date and speed in question, resulting
in one longitudinal tensile strain and one transverse tensile strain for each date, speed, and
60
axle type. Selecting the maximum strain gave the best hit on the gauge array, such that the
strain selected was that induced by the axle group pass that came closest to traveling directly
over the strain gauge. It was imperative that the strains compiled included all six gauges in
each direction as wheel wander was a factor.
Strain vs. Speed
The strains determined from the best hit selections were investigated to determine the
relationship between vehicle speed and tensile strain. For the first two dates, April 6, and
April 10, 2007 the induced longitudinal strain decreases with an increase in speed. The same
trend exists for the last two dates; however, the strain decreased at a greater rate than on the
first two dates for each increase in speed. In looking at the rate of strain change due to
vehicle speed relative to mid-depth temperature it was found that at high mid-depth
temperatures the longitudinal strain was reduced at a greater rate, whereas at colder middepth temperatures, the rate of reduction nears zero. Therefore, at warmer temperatures,
vehicle speed has a greater influence on the tensile strains at the bottom of the HMA than at
colder temperatures. This was to be expected since warmer temperatures induce a greater
viscoelastic behavior from the HMA thus increasing the impact of vehicle speed.
For all test dates and axle types the same general trend existed: increasing the vehicle speed
resulted in a decrease in the amount of strain that was induced. Attaching a trend line to the
data, it became evident that the longitudinal strain was proportional to the natural logarithm
of the speed of the applied load, modeled by Equation 5-1. The goodness of fit,
corresponding to the trend line for each axle type on each test date revealed that Equation 5-1
models the relationship very well, returning R2 values ranging from 0.80 to 0.99 with the
majority of the R2 values greater than 0.90. The same trends existed for strains under each
axle type in the transverse direction. Equation 5-1 can also be used to describe those strains,
returning relatively high R2 values.
(5-1)
b
t a ln v
where:
t = tensile strain (microstrain)
v = speed of applied load (mph)
a, b = regression coefficients
61
Strain vs. Temperature

By conducting the experiment on four different dates throughout the month of April and
beginning of May, a wide range of in-situ pavement temperatures were experienced.
Pavement surface temperatures (oF) ranged from the low 70s to low 120s. Temperatures
from all probes were recorded prior to the start of each test. As to be expected, the
temperature throughout the structure rose as the experiment progressed throughout the month
of April and into the month of May. Additionally, surface temperatures were significantly
higher than temperatures at the bottom of the asphalt concrete. Mid-depth temperatures were
chosen for the analysis of the strain-temperature relationship under each axle type and speed
tested. A previous investigation at the Test Track also selected mid-depth temperature as the
best predictor of strain (33, 29).
To further investigate the effect of temperature on strain, the maximum tensile strain was
plotted against the mid-depth temperature. In general, an increase in mid-depth pavement
temperature resulted in a very large increase in tensile strain, a relationship that can be
described by an exponential function, of the form listed in Equation 5-2. Regression
equations and coefficients of determination (R2) were generated for the strain due to
temperature at all vehicle speeds, returning consistently high R2 values for all four speeds,
ranging from 0.96-0.99. These high R2 values indicate that Equation 5-2 describes the
temperature-strain relationship very well. Furthermore, the speed-strain relationship was
reiterated in this comparison, such that the highest strain values were induced at the slowest
test speeds. Thus, the largest longitudinal strain occurred under a single axle traveling at a
speed of 15 mph, at the highest mid-depth pavement temperature, 99.47 F. These findings
were also evident in the other two axle types: steer and tandem axles, however, the single
axle consistently produced higher strain levels under loading. Similar trends were also found
for the tensile strain induced in the transverse direction. The relationship described by
Equation 5-2 was used to characterize the strain under any of the three axle types and the
strain in either direction (longitudinal or transverse).
(5-2)
cedT
t
where: t = Tensile strain (micro strain)
T = Mid-depth Temperature (F)
e = constant, with approximate value = 2.71828183
c, d = regression coefficients
Combined Effect
It is of some benefit to know the effect of speed and temperature combined, as independently
these variables greatly influence the amount of tensile strain induced in a pavement. The
investigations into the effects of speed and temperature revealed that tensile strain is directly
proportional to the natural logarithm of the vehicle speed and directly proportional to the
exponential function of temperature. To determine the combined effect, DataFit, a regression
analysis software program, was utilized to linearly combine the above equations into
regression equations of the form:
62
f ln v e gT h
where:
t = tensile strain (micro strain)
v = speed (mph)
T = mid-depth pavement temperature (F)
e = constant, with approximate value = 2.71828183
f, g, h = regression coefficients
t
(5-3)
The above equation was selected to represent the effect of speed and temperature based on
the coefficients of determination and the significance of each coefficient. The regression
coefficients for each axle type and gauge orientation are listed in Table 5.1 and the
coefficients of determination were consistently very high for each regression equation.
Although not shown, each coefficient was found to be statistically significant with all pvalues reported as less than 0.001.
TABLE 5.1 Regression Coefficients by Axle Type and Direction of Strain (28)
Gauge Orientation
Longitudinal
Longitudinal
Longitudinal
Transverse
Transverse
Transverse
Axle Type
Steer
Tandem
Single
Steer
Tandem
Single
Regression Coefficients
f
g
h
-24.25 0.047 108.81
-34.67 0.051 172.93
-40.57 0.053 206.67
-19.70 0.050 101.40
-15.05 0.500 112.57
-16.23 0.051 102.18
R
0.955
0.970
0.983
0.986
0.971
0.986
Application of Regression Models

The regression models described above serve two functions. The first is to characterize the
effects of vehicle speed and mid-depth temperature for the pavement section investigated.
The second is to provide a way to predict the tensile strains under a variety of conditions.
Both are beneficial for perpetual pavement design in that the regression models can be used
to determine the conditions that induce strains beyond the laboratory-established threshold
limit. Further analysis stemming from these developed equations can also be utilized to
assess the relevance of laboratory-established thresholds to field conditions.
To determine the most critical conditions for the perpetual pavement investigated, five
speeds (15, 25, 35, 45, and 65 mph) and five mid-depth pavement temperatures (40, 60, 80,
100 and 120 F) were selected to predict tensile strains from the aforementioned regression
equations. The vehicle speeds were representative of those included in the experiment, with
the addition of one speed (65 mph) outside the range that was selected to reflect highway
speeds. For the mid-depth pavement temperatures chosen for this analysis, three
temperatures were experienced in the experiment and the two additional temperatures (40
and 120 F) allowed for extrapolation of strains to temperatures experienced at the Test Track
but not specifically part of this experiment.
The longitudinal tensile strains predicted for a single axle were plotted in Figure 5.1, as well
as a threshold level of 100 . As expected, a vehicle speed of 15 mph is most critical for the
63
N9 test section, inducing strains above the threshold for all of the selected temperatures. The
critical mid-depth pavement temperature was found to be approximately 78F for vehicle
speeds of 65 mph or less. At these conditions the predicted strains were equal to the
threshold value of 100 ; therefore, at lower mid-depth temperatures, zero damage would
theoretically be incurred under an infinite number of load applications.
700
600
500
Microstrain
15 mph
25 mph
400
35 mph
45 mph
300
65 mph
Strain Threshold
200
100
0
20
40
60
80
100
120
140
Temperature
FIGURE 5.1 Predicted Strain by Temperature for a Single Axle.

To investigate the relevance of the laboratory established thresholds, it is important to
consider how frequently the critical conditions occur or are surpassed. As noted earlier,
section N9 was also subjected to routine traffic at 45 mph for 16 hours per day, 5 days a
week, as part of the overall trucking operation for the Test Track. This operation, which
commenced in November, 2006 and continued to the end of the Test Track cycle, provided a
range of temperatures experienced over a much broader time period. Temperatures were
recorded on a minute-by-minute basis from which hourly summaries were stored. From the
available temperature data, the average mid-depth temperatures were retrieved from
November 7, 2006 to January 14, 2008. Using the regression equation developed above, the
tensile strains at the bottom of the HMA were calculated for the recorded temperatures and a
vehicle speed of 45 mph.
Figure 5.2 depicts the calculated longitudinal strain under a single axle for the average middepth pavement temperature data (T2_avg) and vehicle speed of 45 mph for the hours of
operation (5:00AM to 11:00PM). Using the same regression equation for the longitudinal
strain under a single axle, the critical temperature for a vehicle speed of 45 mph was found to
be 72.83F. Figure 5.2 illustrates that on a large number of occasions mid-depth
64
temperatures far surpassed this critical temperature, with maximum mid-depth temperatures
approaching 118F. Corresponding to these high temperatures were elevated tensile strains
well above the strain threshold level of 100 , with maximum strains near 575 . For the
dataset that spanned just over a year, approximately 55% of the calculated strains fell below
the 100 threshold level. To date, this section has exhibited only minimal amounts of
rutting (< 5mm) and no cracking. Given the high strain levels and good performance thus
far, it may suggest that the field strain threshold could be greater than 100 . However,
further traffic and testing is certainly warranted to confirm or refute this statement. The
strain regression equations developed in this study can be used to further study this test
section through the remainder of the trafficking cycle.
140
700
120
600
100
500
80
400
60
300
40
200
20
100
0
10/10/2006
12/24/2006
3/9/2007
5/23/2007
8/6/2007
10/20/2007
1/3/2008
Microstrain
Temperature (oF)
T2_AVG
Strain
0
3/18/2008
Date
FIGURE 5.2 Measured Strain and Mid-Depth Temperatures for the N9 Test Section
(28).
Summary
Investigations at the Test Track helped characterize those effects on pavement response for a
perpetual pavement section, using the N9 test section. Results from this investigation
revealed the following:
The rate of strain reduction was more sensitive to vehicle speeds at warmer
temperatures.
The tensile strain at the bottom of the HMA layer was found to be proportional to the
natural logarithm of the vehicle speed.
The mid-depth pavement temperature correlated best with the induced tensile strain.
Increasing the mid-depth temperature resulted in an exponential increase in tensile
strains. The combined effects were correlated to the measured strain levels, resulting
in a regression equation of the form:
65
(5-4)
f ln v e gT h
This comparison between predicted strains and an established laboratory strain
threshold indicated that the critical mid-depth pavement temperature for vehicle
speeds of 65 mph or less was approximately 78F. Generally speaking, for vehicle
speeds of 65 mph or less, the strains induced at temperatures greater than this critical
temperature will be greater than the threshold level, causing damage to the pavement
that must be accounted for.
For a time period of just over a year it was found that less than 55% of the estimated
strains were below a threshold value of 100 .
Based on these analyses, monitoring of strain levels and performance of section N9
should be continued to determine if a threshold level of 100
is conservative for the
section.
t
FIELD-BASED PERPETUAL PAVEMENT STRAIN THRESHOLDS

M-E pavement design and analysis has recently made advances toward widespread
implementation throughout the United States. As the new MEPDG is being completed and
implemented, more attention is being spent on proper material and pavement response
characterization (34). Material properties are needed in this design framework to determine
theoretical load-induced responses in pavement structures. These responses are then used in
transfer functions to predict the life of the pavement through Miners Hypothesis (35).
Transfer functions rely on theoretical strains and pressures to estimate the design life of
pavement structures. If these values are accurately estimated, the transfer functions will
provide the engineer with a pavement of optimized thickness.
Since perpetual pavement design relies on maintaining pavement responses below some
critical thresholds, it is well suited to M-E pavement design. To capture the fatigue life of
pavements, engineers can estimate pavement responses so the pavement will have sufficient
life. In perpetual pavement design, the determination of fatigue life is controlled by the
tensile strain at the bottom of the HMA. This mechanistic response is limited to mitigate the
possibility for bottom-up fatigue cracking (31; 36; 37). The longitudinal strain at this
pavement location has proven to be critical in thinner pavements, and in a fully-bonded
pavement, it is always the location of highest tensile strain (38).
In a 2006 survey of APT facilities in the United States, 85.7% of the responding facilities
measured horizontal strain at the base of the HMA layer to study fatigue life (39). Other
projects on actual highways, such as I-5 in Oregon and the Marquette Interchange, have also
incorporated measuring strain at the base of the HMA into their research (40). In protecting
a perpetual pavement against fatigue cracking, engineers typically design the pavement so
that the tensile strain at the base of the HMA remains below 70 (37). Other engineers
suggest that the strain threshold be between 60 to 100 based upon laboratory testing (41).
A recent experimental pavement project in China allowed perpetual pavement design to
reach the seemingly unconservative value of 125 (42).
It is important for engineers to determine an appropriate and reliable value to use as a strain
threshold in M-E perpetual pavement design. While many engineers currently use a 70-100
66
fatigue threshold, these values come from laboratory testing at one temperature with no
rest periods. It is difficult, if not impossible, to relate these values and correlate them to field
testing. If the field design threshold could be raised, based on sound research and
engineering, highway agencies would be able to build thinner pavements. A perpetual
pavement experiment in China reduced its thickness from 20 to 15 inches when the fatigue
threshold limit was increased from 70 to 125 (42). This would reduce the construction
and material costs for highway agencies at a time where funding is extremely limited during
these frugal economic conditions. This would also substantially reduce the amount of nonrenewable natural resources consumed for road construction. In the end, the pavement
system as a whole would be more efficient.
Objective and Scope
The main objective of this research was to recommend strain criteria for fatigue cracking in
perpetual pavements. This was completed by correlating simulated and measured strains at
the bottom of the HMA layer to field performance over three cycles of the Test Track. For
the 2000 Track experiment, strain distributions were estimated using PerRoad, a mechanistic
modeling software package, for six sections along the North and South tangents. For the
2003 cycle, direct strain measurements from the base of the HMA were used to develop
strain profiles for seven structural sections. Direct strain measurements for eight sections
from the 2006 Test Track were also analyzed.
2000 Test Track
Six thick (greater than 24 inches) test sections built in the summer of 2000 were selected for
a theoretical strain analysis: N11, N12, S2, S9, S10, and S13. These six sections were chosen
because they had all experienced at least 20 million ESALs without showing signs of fatigue
cracking. Since pavement response instrumentation was not included in the 2000
experiment, strains were estimated using the mechanistic pavement modeling program
PerRoad. Inputs for PerRoad include trafficking data, layer thicknesses, and material
properties of the HMA and soil. Trafficking databases have been kept at the Test Track since
its inception; therefore, the traffic could be divided into a load spectrum for the first 20
million ESALs of the experiment (2000-2005). This traffic division is documented
elsewhere (29).
Once the load spectrum was determined, the exact layer thicknesses needed to be imported
into the program so PerRoad would know where to calculate the desired strains. Each test
section at the 2000 Test Track consisted of a 4 inch experimental surface mix on top of a
perpetual buildup. The perpetual buildup was essentially the same for each section around
the track with slight variations in constructed thickness. The subgrade for the experiment
was an improved roadbed material, also documented as Track Soil, taken from rock
formations in the West curve of the Test Track compacted to 95% of Proctor maximum
density. Above the subgrade was another 12 inch lift of the improved roadbed material;
however, this lift was compacted to 100% of Proctor for a higher density. A densely crushed
granite base layer, often used by ALDOT, was built in a 6 inch lift above the Track soil. The
final layer separating the unbound materials from the bound materials was a non-woven
geotextile fabric. This layer was placed above the granular materials to allow the flow of
67
water through the layer, but not allow fines to pump to the surface or through the pavement
layers (1).
The perpetual buildup of the 2000 Test Track included three layers of bound materials as
well as the previously mentioned unbound materials. The bottom layer of the bound
materials was 4 inches of permeable asphalt treated base (PATB). While four inches of
material was planned as a part of the experiment, the thick spread rate made the build-up
closer to 5 inches in most areas of the Track. 15 inches of Superpave mix was placed on top
of the PATB layer. The bottom nine inches of the Superpave mix consisted of a mix design
using a PG 67 binder. The top six inches used a PG 76 binder to increase the rutting
resistance (1).
After building each sections cross-section, the stiffness of the HMA and the resilient
modulus of the soil could then be characterized. This was accomplished using falling weight
deflectometer data collected during 2001. The AASHTO two-layer backcalculation
methodology was followed to determine the material properties in question on the days of
testing (43). In this analysis, the division between the two layers occurs at the geotextile or
beneath the PATB. Since the subgrade is a stiff material, it was put into one layer with the
12 inches of improved roadbed and 6 inches of granite base. Relationships were then
developed between the HMA modulus and the temperature 10 inches into the HMA during
the time of testing (Table 5.2). These relationships were used to create a cumulative
distribution for the stiffness of the pavement during trafficking.
During FWD testing, temperatures were recorded 10 inches deep in the pavement. With
these recorded data, non-linear regression was used to develop relationships between the
backcalculated material properties and temperature at the time of the testing. Figure 5-3
shows one example of the relationship developed between temperature and the modulus of
the hot-mix asphalt from section S10 on a semi-log plot.
68
1,000,000
HMA Modulus, psi
y = 648817e-0.0066x
R2 = 0.3693
100,000
0
20
40
60
80
100
120
Temperature, F
FIGURE 5.3 HMA and Temperature Relationship for S10 (29).

When comparing power, linear, and exponential relationships, negative exponential
relationships returned consistently higher R2 values. One would expect a negative
relationship between temperature and HMA modulus because as the temperature in a
pavement structure increases, the pavement should become softer and its modulus would be
reduced. Therefore, the equations developed for the six experimental sections follow the
format of Equation 5-5, and Table 5.2 displays each equations constants and R2 value.
E p k1e k2T
(5-5)
Where: Ep = HMA modulus, psi
k1 = section specific constant
k2 = section specific constant
T = temperature, F
TABLE 5.2 Constants for Equation 5-5
Section k1
k2
R2
N11
816061 -0.0087 0.73
N12
797238 -0.0091 0.47
S2
384073 0.0036 0.09
S9
605842 -0.0093 0.71
S10
648817 -0.0066 0.37
S13
747956 -0.0107 0.75
69
As Table 5.2 indicates, three of the sections maintained a relatively strong relationship
(R2>0.7) between temperature and the modulus of the pavement. Two of the sections were
weaker with R2 values ranging from 0.37 to 0.47. Section S2 had no relationship that could
be developed between its HMA modulus and the temperature. The lack of relationship
prevented a multi-seasonal analysis from being conducted on section S2. This was also
verified in laboratory experiments.
Upon the development of these relationships, cumulative distributions of the stiffness were
developed by calculating stiffness based on the average hourly temperature under trafficking.
These cumulative distributions were then divided into quintiles for PerRoads five season
temperature analysis.
PerRoad also required a characterization of the soil in each cross-section. Figure 5.4 shows
the results of comparing S10s resilient modulus of the soil to the pavement temperature
during the FWD testing. As can be seen, there is no real trend upwards or downwards with
temperature for the soils material properties. This was the case for each of the six sections
analyzed from the 2000 Test Track. Since no relationship could be developed, the average
modulus from the FWD testing was used along with the coefficient of variation of the
backcalculated data to simulate the variability of each sections soil modulus.
With both the material properties of the pavement and soil characterized, the load spectrum
accurately quantified, and cross-sections for each test section constructed, the needed dataset
required for PerRoad to complete its mechanistic pavement analysis was complete. The
program used the data created and coupled it with layered elastic theory to calculate
horizontal strain at the base of the bound materials, the critical location for fatigue cracking.
Once these strains had been estimated, cumulative distributions were developed. Figure 5.5
presents the cumulative distributions for all the sections analyzed in the 2000 Test Track.
This plot represents the percentiles of strains experienced at the base of the HMA layer in
each section over six years of trafficking between 2000 and 2005.
As expected, these test sections proved to be overdesigned for fatigue cracking. At the 99th
percentile, all six sections have strains below 9 . The maximum 50th percentile of the six
sections is even below 4.5 . The 99th percentile value is 8 times smaller than the laboratory
estimated fatigue threshold of 70 . The 50th percentile value is 15.5 times lower than the
estimated laboratory fatigue threshold. Because these strains are so much lower than this
estimated fatigue threshold, it is possible to see that pavements can be placed thinner than 23
inches with the possibility of having adequate fatigue protection as long as proper
construction practices and mix design are considered. Exactly how thin was determined by
examining sections in the 2003 and 2006 structural studies.
70
60000
Soil Resilient Modulus, psi
50000
40000
30000
20000
10000
0
0
20
40
60
80
100
120
Temperature, F
FIGURE 5.4 Soil Resilient Modulus and Temperature Relationship for S10 (29).
100%
90%
80%
Percentile
70%
60%
50%
N11
40%
N12
30%
S2
S9
20%
S10
S13
10%
0%
0
Estimated Microstrain
FIGURE 5.5 Cumulative Distributions of Strain for 2000 Test Track (29).
71
10
2003 Test Track
The 2003 Test Track introduced the structural study where embedded asphalt strain gauges
measured strain at the base of the HMA layer of a pavement structure. PerRoad was not
needed since strain levels were directly measured. The two components of the analysis
requiring manipulation and study were the loading configuration and the pavement
responses. Detailed trucking databases allowed precise loading configurations to be analyzed
and weekly measured pavement responses were used to develop continuous strain
distributions for the structural sections. These two design components were then linked to
the observed pavement performance of the section to make correlations between pavement
response and performance in seven of the eight structural sections. N8 was excluded from
analysis due to the debonding that occurred during the 2003 Test Track (11).
The trafficking of the 2003 Test Track was completed using a fleet of four triple flat-bed
trailers and one single box trailer. Previous research was conducted which developed
relationships by section for each truck between a mechanistic pavement response (strain) and
temperature (33). These relationships were used to develop cumulative distributions of strain
for the life of the pavement (i.e. until cracking was observed).
While Priest and Timms work (33) developed strain-temperature relationships for the seven
structural sections in this study, the work did not continue into developing cumulative strain
distributions for each structural section. Figure 5.6 provides the methodology used in
developing the strain distributions for this analysis. These calculations were made by linking
the trafficking database to the strain database to determine the number of times each strain
magnitude occurred for each section.
Once the cumulative distributions were determined for each structural section, the 1st, 99th,
and every 5th percentile were manually picked from the completed spreadsheets to develop
cumulative distribution plots. Figure 5.7 provides the cumulative distribution functions by
section graphically.
Quantify Strains from
Strain-Temperature
Relationships (Priest, 2005)
Determine Repetitions of
Each Microstrain Level
Develop a Cumulative
Distribution Plot for Each
Structural Section
Determine
Percentage of
Repetitions for
Each Microstrain
FIGURE 5.6 Developing Cumulative Strain Distribution Methodology.
72
100%
90%
80%
Percentile
70%
60%
50%
N1 (Cracked)
40%
N2 (Cracked)
N3 (No Cracking)
30%
N4 (No Cracking)
20%
N5 (Cracked)
N6 (Cracked)
10%
N7 (Cracked)
0%
0
200
400
600
800
1000
Microstrain
FIGURE 5.7 Cumulative Distribution Plots for 2003 Test Sections (29).
Of the seven sections analyzed in the 2003 analysis, five of the sections experienced fatigue
cracking: N1, N2, N5, N6, and N7. N3 and N4 have not shown any signs of fatigue cracking
after 20 million ESALs.
Comparing the cumulative distributions of these sections, section N4 has the lowest
distribution strains at the high end of the cumulative distribution. However, until the 55th
percentile strain, its strains are not the lowest. Those are found in N2, but the strains soon
escalate in this thinner section.
The other section that did not show signs of fatigue cracking, N3, started out with low strains
comparable to those in failed sections; however, from the 25th until the 50th percentile, its
strains track very well with sections that failed in fatigue. The 55th percentile seemed to be
the breaking point for sections N3 and N4 where failed sections began to escalate into higher
strain levels.
While the breakpoint where the cumulative distribution functions of the cracked sections are
greater than those of the non-cracked occurs at the 55th percentile, clear deviations between
the sections that performed well and those that did not are present after the 60th percentile
where the difference between the maximum non-cracked sectional strain and minimum
fatigue cracked sectional strain jumps from 4 at the 55th percentile to almost 30 at the
60th percentile.
73
Comparing the cumulative distribution functions calculated for sections N3 and N4 to the
often used laboratory fatigue thresholds of 70 or 100 , one would find that for both
sections, less than 10% of the strain measurements fall below 70 . In section N3, less than
15% of the strain measurements were below 100 , and this value was under 25% for
section N4. Therefore, it can be inferred that measured strains in the field can exceed the
laboratory fatigue threshold without fatigue damage occurring. This will allow for the design
of thinner pavement structures in the future.
These limited data suggested that pavements can be designed to withstand higher magnitude
strains in the field than previously speculated from laboratory testing. Further investigation
of the 2006 structural sections was needed to more precisely quantify acceptable strain levels.
2006 Test Track
In 2006, the Structural Study at the Test Track grew to 11 sections as previously noted in
Chapter 4. Since instrumentation was incorporated into these sections, actual strain
measurements could be taken at the base of the HMA. Only eight sections were included in
the 2006 analysis because sections N5-N7 had already experienced cracking.
The methodology followed to create the cumulative strain distributions for the 2006 Test
Track was similar to that of the 2003 Test Track; however, previous research had not been
done to develop strain-temperature relationships. In 2006, it was decided to develop
relationships between strain and temperature by axle for each section instead of by truck.
When analyzing pavement response by axle, each steer and tandem axle was processed for
every truck pass; however, of the five trailing single axles, only the axle with the best hit
was processed. The best hit was defined as the axle yielding the highest recorded strain.
This was believed to be the axle which most directly hit a strain gauge. The best hit single
axle is circled in Figure 5.8.
Figure 5.8 also shows that the strain trace for the triple trailers predominantly used at the Test
Track consisted of eight individual strain events. These were the results of one steer axle,
one tandem, and five single axles loading the pavement. A double trailer resulted in six
strain events, and a single trailer had four. In each case, no matter how many single axles
were present, only the best hit was processed.
74
Right Gauge
500
Longtidinal Microstrain
400
300
200
100
0
-100
-200
0.5
0.7
0.9
1.1
1.3
1.5
1.7
1.9
2.1
2.3
2.5
Time, sec
FIGURE 5.8 Best Hit Single Axle (29).

The 2003 Test Track strain-temperature relationships were developed using only longitudinal
strains. When analyzing the data, longitudinal strains were typically larger than transverse
strains at the same lateral offset, and a previous study (44) had shown longitudinal gauges
were less influenced by wander and would better represent what is occurring at the base of an
HMA layer under trafficking. Hence, longitudinal strains were used to develop straintemperature relationships for the 2006 Test Track.
The next decision as to which strains to use in developing the strain-temperature
relationships was what magnitude strain to incorporate in the analysis. While great care was
taken to ensure only quality data were included in the strain databases, voltage spikes, faulty
gauges, and processing errors occur in data processing. If the maximum longitudinal strain
were chosen to develop the strain-temperature relationships, these errors might drive the
strain temperature relationship development. Upon a thorough examination of the data, it
was determined the 95th percentile strain would be appropriate strain magnitude to use for
relationship development.
Previous researchers at the Test Track (33) developed strain-temperature relationships based
on the mid-depth temperature. To remain consistent with the previous research, the middepth temperature was once again used to develop new strain-temperature relationships. The
mid-depth temperature was directly measured in the newly constructed sections (N1, N2, N8,
N9, N10, and S11), and linear interpolation was used to determine the mid-depth temperature
for the sections remaining from the 2003 experiment (N3 and N4).
75
Figure 5.9 provides the methodology for developing the strain-temperature relationship.
Queries were first developed in Microsoft Access which matched the 95th percentile strain for
each axle and section to the temperature measured at the time of testing recorded in another
database. These matched records were then exported to Excel where non-linear regression
was used to determine the best relationship between the 95th percentile longitudinal strain and
the mid-depth temperature.
Determine 95th
Percentile Strain for
Single, Tandem, and
Steer Axles by Section
Determine Mid-Depth
Temperature for Each Section
at Time of Testing
Microsoft
Access/Excel
Steer Axle
StrainTemperature
Relationship
Tandem Axle
StrainTemperature
Relationship
Single Axle
StrainTemperature
Relationship
FIGURE 5.9 Relationship Development Methodology.

To determine the best fit equation, Excel was used to fit power, exponential and polynomial
functions to the data. The exponential equations returned the most consistently high R2
values. Therefore, equation 5-6 was used as the backbone for the strain-temperature
relationships in 2006.
(5-6)
k1e k2T
Where: = strain, microstrain
k1 = sectional constant
k2 = sectional constant
T = temperature, F
An example of a relationship developed for the steer axles is graphically provided in Figure
5.10. Table 5.3 provides an example of the relationship coefficients for steer axles. The
other coefficients are documented by Willis elsewhere (29).
76
When looking at the coefficients, the thicker sections (N8 and N9) had the lower k1 values in
all three analyses. At lower thicknesses, the k1 increased somewhat with the exception of
S11 which was thinner than N3 and N4; however, these two sections have oxidized and
become stiffer over the course of the previous trafficking cycle. The k2 coefficients in the
equation give the slope or rate of strain increase with temperature.
As can be seen in the Table 5.3, section N4 has a high k1 coefficient; however, the low k2
coefficient means that strain does not increase much with temperature. The opposite is seen
in section N9. The k1 value is lower than any other section, but it has the one of the highest
k2 values. Both of these sections were free of fatigue cracking. The two sections with the
most fatigue cracking (N10 and S11) had comparatively high k1 and k2 coefficients; thus, as
temperatures increase in these two sections, the strain magnitudes will escalate greatly.
250
Microstrain
200
150
y = 6.6066e0.0284x
R2 = 0.9107
100
50
0
0
20
40
60
80
100
120
Temperature, F
FIGURE 5.10 Strain-Temperature Relationship for Steer Axles in N9 (29).

After these relationships had been developed, the previously mentioned methodology was
followed for linking the strain magnitudes to trafficking repetitions for the development of a
cumulative distribution function. The cumulative strain distributions for the 2006 sections
are provided in Figure 5.11.
77
TABLE 5.3 Equation Coefficients by Section for Steer Axles
Section k1
t-stat k1 k2
t-stat k2 R2
N1
21.58 7.30
0.0272 -3.82
0.84
N2
24.605 7.54
0.0264 -3.45
0.81
N3
10.543 13.15
0.0316 -7.51
0.80
N4
17.631 12.10
0.0225 -5.16
0.79
N8
12.521 12.07
0.035 -7.51
0.87
N9
6.6066 15.77
0.0284 -8.20
0.91
N10
20.677 11.10
0.0323 -6.62
0.69
S11
14.607 15.87
0.037 -10.02
0.75
100%
90%
80%
Percentile
70%
N1 (Top-down)
60%
N2 (Top-down)
50%
N3 (No crackng)
N4 (No cracking)
40%
N8 (Cracked)
N9 (No cracking)
30%
N10 (Cracked)
S11 (Cracked)
20%
10%
0%
0
200
400
600
800
Microstrain
1000
1200
1400
FIGURE 5.11 Cumulative Distribution of Strains for 2006 Structural Sections (29).
Three of the sections provided in Figure 5.11 experienced fatigue cracking: N8, N10, and
S11. Previously-mentioned research showed N1 and N2 also cracked, but the cracking in
these sections was top-down and not bottom-up fatigue cracking.
If one were to compare the cracked profiles versus the non-fatigue cracked profiles, a clear
breakpoint occurs at the 45th percentile. At this point, the cracked sections begin to diverge
greatly from the non-cracked sections. N3 and N4s strain profiles represent the least
conservative strain profiles that were able to withstand trafficking without fatigue cracking.
These two sections have also received double the traffic of any other section in the 2006
study. Therefore, these two sections should be considered when determining the uppermost
bound for a field-based measured fatigue threshold.
78
Threshold Development
There are three distinct groups of cumulative strain distributions from the sixteen sections in
this investigation (Figure 5.12). The first group is the sections that were overdesigned in
2000. Pavements do not need to be designed at this thickness to prevent fatigue cracking.
The second and third groups break apart from each other about the 55th percentile. The group
with the smaller magnitude strains includes the sections which did not experience fatigue
cracking. The group with the larger strains did experience fatigue cracking. Therefore, a
distinction can be made between the cumulative strain distributions of the sections that failed
in fatigue to those that did not.
100%
N1 2003
90%
N2 2003
80%
N3
N4
70%
Percentile
N5 2003
60%
N6 2003
N7 2003
50%
N1 2006
40%
N2 2006
N8 2006
30%
N9 2006
20%
N10 2006
S11 2006
10%
S13 2000
0%
0
200
400
600
800
1000
1200
Microstrain
FIGURE 5.12 Cumulative Distributions from Three Test Cycles (29).

Three criteria were considered to help develop a new strain-criterion for flexible perpetual
pavement design. (1) The section could not be overdesigned. (2) The section could not have
exhibited any fatigue cracking. (3) They section had to have experienced at least 20 million
ESALs. Four analyses met these criteria: N3 by axle, N4 by axle, N3 by truck, and N4 by
truck. When these four analyses were carefully analyzed, it was discovered that they were
not very different from each other numerically. Previous research conducted at the Test
Track found that duplicate strain gauges should typically be within 30 microstrain (15 )
of each other. If an average of all four test sections was taken and a 15 confidence
boundary was put around that average (Figure 5.13), all four cumulative strain distributions
would fall within that confidence boundary. Therefore, the average of the aforementioned
analyses was determined to be an appropriate field-based strain threshold for flexible
perpetual pavement design. These values are given in Table 5.4. If these values are
79
validated for designing perpetual pavements with similar stiffness characteristics and similar
loading scenarios, then pavements can be designed thinner without losing performance life.
100%
90%
80%
Percentile
70%
60%
Average
50%
+15
40%
-15
N3
30%
N4
20%
10%
0%
0
50
100
150
200
250
300
350
400
450
Microstrain
FIGURE 5.13 Average Strain Distribution with Confidence Bands (29).

Conclusions
The following conclusions were drawn from the previously discussed research:
Fatigue cracked sections had wider strain profiles when compared to the strain
distributions of uncracked sections.
Higher magnitude strains govern fatigue cracking since many of the cracked and
uncracked strain profiles were similar in the lower percentiles.
Assuming similar traffic and fatigue characteristics, the provided field-based strain
profile (Table 5.4) should withstand fatigue cracking. Implementation of these
findings beyond the scope of the Test Track requires further validation.
80
TABLE 5.4 Field-Based Strain Criteria for Perpetual Pavements (29).
Percentile
99%
95%
90%
85%
80%
75%
70%
65%
60%
55%
50%
45%
40%
35%
30%
25%
20%
15%
10%
5%
1%
Fatigue Limit,
394
346
310
282
263
247
232
218
205
193
181
168
155
143
132
122
116
101
89
72
49.
81
CHAPTER 6 - LABORATORY AND FIELD COMPARISONS
One of the most complex and challenging tasks undertaken at the NCAT Test Track is the
development of relationships between laboratory test results and field measured data or
performance. It is difficult, at best, to capture the complexities that HMA experiences in the
field and translate them into a simple laboratory experiment. Three studies were undertaken
as a part of the 2006 Test Track to relate laboratory test results to field measured or modeled
data. The first experiment characterized the modulus values of the granular materials at the
Test Track and compared them to laboratory resilient modulus results. The second study
compared the dynamic modulus values from the structural test sections and compared them
to models that had been developed by researchers to find a best fit model for E*. The final
analysis took the cumulative strain distributions described in the previous chapter and
compared them to laboratory determined fatigue thresholds.
GRANULAR MATERIALS LABORATORY AND FIELD COMPARISONS
One of the critical inputs for accurate M-E pavement design is accurate characterization of
the stiffness of the unbound pavement material layers. This stiffness is quantified as resilient
modulus, and this value can be determined either through laboratory testing with the triaxial
apparatus or though non-destructive testing in the field with the FWD. Resilient modulus is
typically expressed as a function of unbound material stress-state using a non-linear stresssensitivity model.
To effectively characterize the stiffness behavior of unbound materials, several factors must
be considered. First, does laboratory or field resilient modulus testing provide a better
representation of material behavior? Secondly, is the material stress-sensitive and, if so,
which stress-sensitivity model best quantifies the behavior of that material? Answering these
questions allow for the most accurate quantification of resilient modulus for pavement design
and more accurate modeling of pavement design life.
The overall goal of this investigation was to mechanistically characterize the five unbound
materials utilized in the eleven structural sections at the NCAT Test Track for effective use
in pavement design. Specific objectives included:
A comparison of laboratory and field derived resilient moduli.
An evaluation of common non-linear stress-sensitivity models with respect to
laboratory and field moduli.
Developing a recommendation as to the effective use of unbound material moduli in
pavement design and analysis.
Unbound Materials Used in the Structural Study
The unbound materials used in this analysis were those incorporated into the design of the
2006 structural study. Table 6.1 shows the material gradations for each of the unbound
materials. Note that the three base layer materials (the limerock, granite, and Type 5) have
reasonably well-graded particle distributions, with relatively small percentages of material
retained above the sieve. These materials also exhibit lower amounts of material passing
the #200 sieve than the two materials that are predominantly used as subgrade materials (the
82
Seale and Track soil). The Seale material is a very fine-gained soil, with almost 58 percent
of the material passing a #200 sieve. The Track soil is more gap-graded, with 17 percent of
the material retained on the 1 sieve and 48 percent of the material passing the #200 sieve.
TABLE 6.1 Unbound Material Gradations (27).
Material
Limerock
Seale
Type 5
Layers
Base
Subgrade
Base
Sections
N1, N2
N8, N9
N10
Sieve
1 1/2"
1"
3/4"
1/2"
3/8"
#4
#8
#16
#30
#50
#100
#200
100
100
100
88
81
61
44
32
26
23
21
18.8
Track Soil
Subgrade
All (N8, N9
Base)
Percent Passing Control Sieve
100
100
100
100
99
83
100
97
81
100
92
78
100
88
75
100
79
71
100
71
68
99
64
66
98
58
64
92
49
61
82
36
56
57.7
25.1
48.0
Granite
Base
S11
100
95
88
83
78
57
47
39
31
23
15
10.2
Laboratory Testing
For this study, triaxial resilient modulus testing was performed in accordance with NCHRP
1-28A Laboratory Determination of Resilient Modulus for Flexible Pavement Design.
This testing was performed on each of the five unbound materials present in the eleven
structural sections at the Track. This testing was subcontracted to Burns, Cooley, Dennis,
Inc. The stresses under which the material was tested is based on material type and whether
or not that material is used primarily as a base/subbase or a subgrade material, and three
replicates of each material type were tested in accordance with the NCHRP 1-28A procedure.
For each unbound material tested, stress and modulus data generated from testing the three
samples were combined into one database for each material. For each testing sequence, the
following data were generated: the total axial stress (1), the confining pressure (2 and 3),
and the calculated resilient modulus. Given these stresses, the bulk stress (), deviatoric
stress (), and octahedral shear stress (oct) were calculated for each testing sequence using
equations 6-1, 6-2, and 6-3, respectively.
For this study, four common non-linear models relating the material stress-state to the
resilient modulus were evaluated for each unbound material. The regression coefficients for
these models were generated for each unbound material by entering in the stress-state and
resilient modulus data into DATAFIT, a non-linear regression modeling software package
developed by Oakdale Engineering. This software is capable of generating multiple non-
83
linear models from a given data set containing one or more independent variables and one
dependent variable. This software also allows the user to define the model for which the
software will calculate the regression coefficients (k1, k2, etc.) as well as pertinent statistics
such as those regarding model fit (R2) and statistical significance of the calculated regression
coefficients (p-values).
1
d
oct
1
(
3
(6-1)
3
1
)2 (
)2 (
)2
(6-2)
(6-3)
Where: = Bulk Stress, psi

d = Deviatoric Stress, psi
oct = Octahedral Shear Stress, psi
1 = Axial Stress, psi
2, 3 = Confining Stress, psi
Equations 6-4 and 6-5 are single-variable stress-sensitivity models that relate resilient
modulus to bulk stress and deviatoric stress, respectively. These models are commonly
specified based on whether the material is a coarse-grained (Equation 6-4) or fine-grained
soil (Equation 6-5). Equations 6-6 and 6-7 are multi-variable stress-sensitivity models that
model resilient modulus as a function of two stress terms. These models are more universal
given they are not constrained for use with a particular soil type. In each model, the first
term (the bulk stress) models stress-sensitivity as a function of confining pressure while the
second term (either deviatoric or octahedral shear stress) models stress-sensitivity due to
shearing stresses. For the purposes of this analysis, equation 6-4 is referred to as the bulk
model, equation 6-5 is referred to as the deviatoric model, equation 6-6 is referred to as the
MEPDG model, and equation 6-7 is referred to as the universal model.
k2
Mr
k1 *
(6-4)
pa
k2
Mr
k1 *
(6-5)
pa
k3
k2
Mr
k1 pa *
pa
oct
pa
k2
Mr
k1 p a *
pa
(6-6)
k3
(6-7)
pa
84
where: Mr = Resilient Modulus, psi

Pa = Atmospheric Pressure (14.7 psi)
k1, k2, k3 = Regression Coefficients
Table 6.2 summarizes the laboratory generated stress-sensitivity models that were generated
based on the laboratory triaxial data from this project. The multi-variable stress-sensitivity
models (the MEPDG and universal models) exhibited much higher model R2 values
(typically above 0.9) than the single-variable stress-sensitivity models (the bulk and
deviatoric models) for the different unbound materials. Both the MEPDG and universal
constitutive models exhibit a good model fit for each of the unbound materials except for the
Track soil material. The Track soil material was clearly the least stress dependent material
tested, with none of the four stress-sensitivity models generating an R2 above 0.7 for this
material. For the laboratory data, an average Track soil modulus of 28,335 psi with a
standard deviation of 6,650 psi was calculated.
To quantify the relative stiffness of the different unbound materials, three representative
stress states for each material were entered into the laboratory generated constitutive
equations (for the MEPDG and universal models). These stress-states were generated by
analyzing the cumulative distribution functions (CDF) of the stress-states used for testing
with each of the different unbound materials. More details regarding the selection of these
representative stress-states are documented elsewhere (27). Table 6.3 lists these
representative stress-states. Figure 6.1 shows the calculated modulus values for each of the
unbound materials at the various stress-states with the laboratory calibrated universal stresssensitivity models.
The granular base materials from the Test Track (Florida limerock, granite base, Type 5
base) exhibited stress-hardening behavior in the laboratory under increasing load while the
Seale subgrade material exhibited stress-softening behavior. The Track soil material showed
no significant stress-sensitivity in the laboratory across different stress-states. The MEPDG
and universal model R2 values of 0.42 and 0.66, respectively, confirm this assessment. In
comparing the materials at equivalent stress-states, the Track soil material exhibited the
highest modulus in the lowest and middle representative stress-states. The limerock base
material was the stiffest material at the highest representative stress-states.
85
Table 6.2 Summary of Laboratory Generated Stress-Sensitivity Models (27)
Model
Form
Material
Limerock
Base
Vulcan
Granite
Seale
Bulk
Subgrade
Type 5
Base
Track
Soil
Limerock
Base
Vulcan
Granite
Seale
Deviatoric
Subgrade
Type 5
Base
Track
Soil
Limerock
Base
Vulcan
Granite
Seale
MEPDG
Subgrade
Type 5
Base
Track
Soil
Limerock
Base
Vulcan
Granite
Seale
Universal
Subgrade
Type 5
Base
Track
Soil
k1
22966.7
p-value
(k1)
0
10862.1
k2
0.4773
p-value
(k2)
0.0000
0.6267
6009.8
14049.7
R2
N/A
p-value
(k3)
N/A
0.5618
0.0000
N/A
N/A
0.886
-0.1201
0.0000
N/A
N/A
0.0288
0.6710
0.0000
N/A
N/A
0.8721
26833.3
0.0447
0.2312
N/A
N/A
0.0179
39001.4
0.2174
0.0000
N/A
N/A
0.2204
21350.0
0.3866
0.0000
N/A
N/A
0.5765
4305.8
-0.5571
0.0000
N/A
N/A
0.7834
29487.2
0.3876
0.0000
N/A
N/A
0.5334
28878.9
-0.0572
0.0465
N/A
N/A
0.0478
1266.8
1.2081
0.0000
-1.2332
0.9326
716.3
0.8468
0.0000
-0.4632
0.9253
817.6
0.3305
0.0000
-3.3946
0.957
883.5
1.0050
0.0000
-0.6575
0.9478
1879.0
0.4067
0.0000
-0.7897
0.4202
717.0
1.2338
0.0000
-0.5645
0.8562
581.1
0.8529
0.0000
-0.1870
0.00001
0.9172
225.1
0.3598
0.0000
-0.7551
0.9786
643.7
1.0318
0.0000
-0.2833
0.9349
1095.4
0.5930
0.0000
-0.4728
0.6642
86
k3
TABLE 6.3 Representative Stress-States used for Modulus Normalization (27)
Bulk Stress
Deviatoric Stress
Octahedral Shear
(psi)
(psi)
(psi)
State 1
10
2
1
State 2
25
7
3.5
State 3
50
15
8
Unbound Material Moduil at Representative Stress-States Universal Model
50000
State 1
State 2
State 3
45000
Resilient Modulus (psi)
40000
35000
30000
25000
20000
15000
10000
5000
0
Limerock Base Granite Base
Seale
Subgrade
Type 5 Base
Track Soil
Material Type
FIGURE 6.1 Unbound Material Moduli at Representative Stress-States Universal

Model (27).
Backcalculation Cross-Section Investigation
The objective of this portion of the study was to generate the optimal cross-section for
backcalculation for each of the eleven structural sections at the Test Track. Optimizing the
backcalculation cross-section will allow for the generation of accurate pavement layer
moduli for each of the structural sections. Confidence in this data set is necessary for the
calibration of viable stress-sensitivity models for the unbound pavement layer materials in
the field.
To generate accurate in-situ moduli for each of the structural sections at the NCAT Test
Track, an investigation was conducted on each of the structural sections to determine the
optimum cross-section for backcalculation. For each structural section, multiple possible
backcalculation cross-sections were generated based on how each section was constructed
and the composition of the different pavement layers. These cross-sections included a
number of 3-layer, 4-layer, and 5-layer pavement systems. Trial cross-sections including a
stiff layer were also included to ensure that bedrock or a shallow water table were not
87
influencing the results of the backcalculation. An example of the trial cross-sections utilized
for sections N1 and N2 are shown in Figure 6.2. A similar set of trial cross-sections was
developed for each of the eleven structural sections based on the construction information
from these sections (27).
Trial Backcalculation Cross-Sections (N1-N2)

(1)-3layer
HMA
(2)-3layer
(3)-3layer
(1)-4layer
HMA
HMA
HMA
HMA
HMA
HMA
HMA
GB
GB
GB
GB
GB
GB
GB +
New Fill
New Fill +
Old Fill +
Subgrade
GB +
New Fill +
Old Fill
New Fill +
Old Fill
(2)-4layer
New Fill
(1)-5layer
(B)-3layer
(B)-4layer
New Fill
New Fill +
Old Fill
Subgrade
New Fill +
Old Fill
Old Fill
Old Fill +
Subgrade
Old Fill +
Subgrade
Subgrade
Subgrade
Subgrade
Subgrade
Bedrock
Bedrock
FIGURE 6.2 Trial Backcalculation Cross-Sections for Sections N1 and N2 (27).

To determine the optimal backcalculation cross-section for each structural section, several
sources of information were utilized. First, FWD testing was performed on four different
dates for each of the structural sections at the Test Track (these data were collected as part of
the regular NCAT FWD testing program). The dates used for this investigation were
selected to encompass a wide variety of pavement temperatures. The pavement temperatures
on these dates ranged from approximately 45 F to over 130 F. For each structural section, the
deflection data from these four dates were backcalculated using each of the trial
backcalculation cross-sections. From this, a database of summary files detailing the layer
moduli and RMS error for each drop was compiled. These individual databases were then
analyzed to determine the quality of the solution.
First, the results from each cross-section were analyzed to determine the percentage of drops
exhibiting a reasonable match between measured and calculated deflections (RMS Error).
For the purposes of this investigation, the percentage of drops that exhibited an RMS error
below 4% was used to determine whether a trial cross-section presented a reasonable
solution. The reasoning behind utilizing 4% RMS as a cut-off value is documented
elsewhere (27). Secondly, the summary files were analyzed to ensure that the solution for a
particular cross-section was stable and presented consistent modulus values between drops at
the same load levels. Erratic layer moduli on subsequent drops at identical load levels serve
as an excellent indicator of solution instability. Additionally, these files were analyzed to
ensure that the modulus values for two adjoining layers were not compensating for one
another on consecutive drops. Finally, the individual solutions were analyzed to ensure that
the modulus values were reasonable for that particular material. The process outlined above
88
typically eliminated several of the trial cross-sections from consideration, leaving only two or
three viable solutions.
The second data set used in determining the optimum backcalculation cross-section was the
data from the FWD on gauge testing conducted on 7/17/07. Precise details of this testing
procedure are documented by Taylor (27). For each structural section, the deflection files
from this testing were backcalculated using the trial cross-sections that were deemed viable
after analyzing the data from the multiple FWD testing dates. These files were
backcalculated to determine the pavement responses at the locations in which
instrumentation was embedded within the pavement structure (e.g. the bottom of the HMA
layer, the surface of the base layer, and the surface of the fill layer). These predicted
pavement responses generated through backcalculation were then paired with the
corresponding measured pavement response from the instrumentation for that particular
FWD loading (drop). Comparing the measured versus predicted strain behavior for the
remaining trial cross-sections were used as a deciding factor in determining the best crosssection for analysis. Measured versus predicted pressures at the surface of the base and fill
layers were also calculated and analyzed to validate the use of the selected backcalculation
cross-section.
Figure 6.3 shows a summary diagram of the final backcalculation cross-sections selected for
each of the structural sections. Given these cross-sections, backcalculation of the deflection
data at the Test Track could then be performed with the aim of characterizing the various
unbound material moduli. Through this investigation, it was shown that bedrock or stiff
layer effects do not seem to influence the deflection data at the Test Track. It was also
shown that 3-layer pavement systems perform much better than 4-layer or 5-layer systems
(possibly due to material similarities between base and fill layers). Full details regarding the
backcalculation cross-section investigation for each of the eleven structural sections at the
2006 Test Track are documented elsewhere (27).
89
N1-N7
N8-N9
N10
S11
0
HMA
HMA
HMA
5
GB
10
GB +
New Fill (1)
GB
Depth (in)
15
GB +
New Fill (1+2) +
Old Fill
20
25
30
HMA
New Fill (1+2) +

Old Fill +
Track Subgrade
New Fill (2) +

Old Fill +
Track Subgrade
Seale Subgrade
+ Track Subgrade
35
40
45
Track
Subgrade
50
FIGURE 6.3 Final Selected Backcalculation Cross-Sections (27).

Field Characterization of Unbound Materials
Methodology
For this study, deflection data collected from each of the structural sections at the Test Track
were used to evaluate the stress-sensitivity of the unbound materials within those sections.
The deflection data used for this study were collected over multiple dates in which a wide
range of pavement temperatures was observed to capture the seasonal variability of these
materials. Backcalculation was performed on the deflection data for each section using the
optimal backcalculation cross-section generated for that section (see previous section). For
each FWD test, this process yielded both a backcalculated modulus value for each of the
pavement layers as well as a comprehensive stress-state at a pre-defined location within each
of the pavement layers. These layered-elastic generated stresses were then adjusted for
overburden so that the true stress-state within the pavement cross-section was represented.
Given this information, four commonly used stress-sensitivity models (listed above in the
laboratory testing section of this report) were generated (using DATAFIT) and evaluated
(through a model calibration and validation process) for the various unbound materials.
These stress-sensitivity models were calibrated based on data from four testing dates that
encompassed a wide range of pavement temperatures and consequently wide ranging stressstates. These calibrated models were evaluated based on model-fit, the statistical
significance of the regression coefficients, and on backcalculated versus predicted modulus
behavior. If the selected models for each material were deemed viable, they next underwent
the process of validation. This process utilized data from four other FWD testing dates (over
a similar wide pavement temperature range) to generate a second database of representative
90
pavement layer stresses and material moduli. These data were then used to determine how
well the calibrated models could predict the unbound material moduli from a different data
set. Additionally, the behavior of the HMA layer moduli with changing temperatures was
evaluated to ensure that the backcalculated solutions were reasonable for each of the test
sections. Full details regarding this investigation are given by Taylor (27).
Model Generation
Figure 6.4 shows the model R2 summary for the various base material types. This figure
shows that the universal model provides the highest model R2 for the field stress and
modulus data. Also, this figure illustrates the superiority of the multiple-variable models
versus the single-variable models in terms of model fit for the base layer materials. The
universal model seems to provide the best fit for both the limerock and the granite base
materials with respect to model R2. This model generates R2 values of 0.83 and 0.67,
respectively, for the two material types. These models were deemed high quality considering
the vast amount of data generated across four days of testing in addition to the large amount
of spatial and seasonal variability generated within the data set. Figure 6.4 also illustrates the
poor model fit exhibited by all four models for the Track soil base and combined Type 5 base
and Track fill materials. This was expected given the relative stress-insensitivity of the
Track soil material in laboratory testing. The combined granite and Track fill in section S11
seems to generate approximately the same model R2 regardless of model used. Table 6.4
summarizes the field-calibrated universal stress-sensitivity models for each of the materials
used at the Test Track. It can be seen from this table that all of the models exhibit
statistically significant regression coefficients except for the S11 model, in which no tangible
benefit was seen in model R2 for the multi-variable models over the single-variable models.
A separate model for the granite base material that excluded sections N5, N6, and N7 was
generated due to there being a large amount of pavement distress in those sections.
Validation of this assumption is shown in Taylor (27).
Figure 6.5 shows a model R2 summary by section and material type for the subgrade layer
materials. To generate a material-specific model for the Track soil subgrade material, several
combinations of the data sets were utilized to generate the best model fit. This was done
because the different cross-sections at the Test Track have this material at various depths
with widely varying materials above the subgrade layer. This figure shows that the sections
featuring only the Track soil material in the subgrade do not exhibit very strong stresssensitivity. Many of these sections exhibit model R2 below 0.5. Additionally, the best R2 for
the material-specific Track soil model (containing the data from sections N1 through N4) is
only 0.44. This result was expected given the fact that this material showed the lowest
laboratory stress-sensitivity of any of the unbound materials. The most appropriate model for
modeling the combined Seale and Track subgrade in terms of model R2 was the universal
stress-sensitivity model based on R2 data and statistical significance of regression
coefficients.
91
R-Squared
R-Squared Comparison - All Model Types - By Material

Type - Base
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Bulk
Dev
MEPDG
Universal
Limerock (N1Vulcan
N2)
Granite (N3N4)
Vulcan
Granite (N3N7)
Track Soil
(N8-N9)
Material Tested
Type 5 Base
Vulcan
+ High
Granite + All
Density Track
Track Fill
Fill (N10)
(S11)
Figure 6.4 Model R-Squared Summary (Base Layer, by Material) (27).

Table 6.4 Field-Calibrated Universal Models (Base Layer, by Material) (27)
pp-value
p-value
Base
k1
k2
k3
value
(k1)
(k2)
Material
(k3)
Limerock
4621.71
0
-2.0788
0
1.8908
0
(N1-N2)
Granite Base
10465.42
0
-3.0449
0
2.7613
0
(N3-N4)
Granite Base
14380.51
0
-3.3340
0
3.0202
0
(N3-N7)
Track Soil
576.16 0.00001 -0.6687 0.00026
0.6202 0.0004
(N8-N9)
Type 5 Base
+ High
Density
8243.90 0.00251 -3.1639
0
2.8650
0
Track Fill
(N10)
Granite Base
+ All Track
769.73
0
-0.1598 0.17916 -0.3246 0.0008
Fill (S11)
92
R2
0.8266
0.6742
0.5603
0.0237
0.2542
0.5244
R-Squared Comparison - All Model Types - By Material Type/Section - Subgrade

1
0.9
Bulk
Dev
MEPDG
Universal
0.8
Model R-Squared
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
N1-N7, N10, S11
N1-N7, N10
N1-N7
Track Subgrade Track Subgrade Track Subgrade
N1-N4
N1-N2
Track Subgrade Track Subgrade
N8-N9
Seale/Track
Subgrade
FIGURE 6.5 Model R-Squared Summary (Subgrade Layer, by Material) (27).

Model Calibration
The next step in the model generation process was to evaluate the measured versus predicted
resilient modulus values for each base material type using the most appropriate constitutive
model. The models that exhibited reasonable R2 values, statistically significant regression
coefficients, and good agreement between measured and predicted resilient moduli were
deemed acceptable. The materials with models that did not meet the criteria of acceptability
were deemed non stress-sensitive. An average and standard deviation of the field-calculated
moduli for these materials were reported in lieu of a stress-sensitivity model.
For the limerock base material (sections N1 and N2), Figure 6.4 shows that the universal
constitutive model provided the highest model R2 for this data set. Table 6.4 shows that each
of the regression coefficients for this model was statistically significant (p-values less than
0.05). Next, the actual backcalculated moduli from this data set were plotted against the
modulus values predicted using the field-calibrated model as illustrated in Figure 6.6. This
figure shows that the data points for this model seem to track closely to the line of unity
(representing the condition where the measured moduli equal the model predicted moduli).
As a result, the universal model was deemed suitable for modeling the stress-sensitivity of
the limerock base material in the field. A similar procedure was carried out for each of the
base and subgrade materials utilized in the structural study at the Test Track.
93
Limerock Base (N1-N2) Universal Stress-Dependency Model Calibration
Predicted Mr (Universal Model) (psi) .
50000
45000
Line of Unity
40000
35000
30000
25000
20000
15000
10000
5000
0
0
10000
20000
30000
40000
50000
Measured Mr (psi)
FIGURE 6.6 Limerock Base Material Universal Model Calibration Data (27).
Model Validation
The next phase in assessing the quality of the field-calibrated stress-sensitivity models was to
evaluate the ability of the model at predicting the backcalculated moduli from a different set
of deflection data. This process of model validation was performed by first compiling a
database containing backcalculated layer moduli and representative stress-states (layeredelastic stresses due to load plus overburden) from four different dates of FWD testing at the
Test Track. This testing was performed on all eleven structural sections over a similar wide
range of pavement temperatures that were experienced during the dates of calibration testing
(45 to 130 F). To validate the generated models, the appropriate stress-states generated
within the various unbound materials under a given FWD load were entered into the
calibrated constitutive equation for that material to generate a predicted layer modulus under
that loading. This predicted modulus was then compared to the measured (or backcalculated)
modulus to assess how well the calibrated equation could predict the backcalculated moduli
for a different data set.
Figure 6.7 shows the validation data set for the universal stress-sensitivity model calibrated
for the limerock base material (shown in Table 6.4). This data set shows a plot of measured
versus model predicted backcalculated moduli using a new set of backcalculation data. The
figure shows that the equation tends to over-predict modulus values that fall below
approximately 5,000 to 7,000 psi. Also, the model tends to under-predict the moduli that fall
above approximately 13,000 to 15,000 psi. This trend can also be seen when the residual
94
values (the difference in measured and predicted moduli) are plotted against the measured
moduli. This analysis is shown in Figure 6.8. Even though the equation over-predicts at low
modulus values and over-predicts at the higher ones, it appears that the vast majority of the
data set fall within 3,000 psi of the measured value. Also, the measured moduli are on the
same order of magnitude as the predicted moduli. In summary, the field-calibrated universal
model for the limerock material offers reasonable, though not ideal, predicted field modulus
values. For future work, perhaps a larger data set for model calibration and validation would
yield a more robust model.
Predicted Mr (Universal Model)

(psi)
Limerock Base (N1-N2) Universal Stress-Dependency Model Validation

50000
45000
40000
35000
30000
25000
20000
15000
10000
5000
0
0
10000
20000
30000
40000
50000
Measured Mr (psi)
FIGURE 6.7 Limerock Base Universal Stress-Sensitivity Model Validation (27).

Limerock Universal Model Residuals
Measured - Predicted Mr (psi)
15000
10000
5000
0
0
10000
20000
30000
40000
50000
-5000
-10000
Measured Mr (psi)
FIGURE 6.8 Limerock Base Universal Stress-Sensitivity Model Residuals (27).
95
Summary of Field-Calibrated Stress-Sensitivity Models
For sections N1 and N2, the universal stress-sensitivity model shown as Equation 8 was
calibrated to model the resilient modulus behavior of the limerock base material under
changing stress-conditions. This model exhibited a very high R2 (0.83) given its calibration
with deflection data taken over a wide range of testing locations and seasonal conditions.
The validation of this model with a different data set yielded mixed results. While the model
seemed to offer a reasonable modulus prediction for the vast majority of the data set, it would
tend to over-predict modulus values below 7,000 psi and over-predict modulus values above
13,000 psi. Therefore, Equation 6-8 is reasonable for moduli in this range, but is not ideal.
A more robust model would likely require a larger data set for calibration and validation.
1.8908
2.0788
Mr
4621.71*14.7 *
14.7
(6-8)
14.7
For sections N3 and N4, the universal stress-sensitivity model shown as Equation 9 was
calibrated to model the resilient modulus behavior of the granite base under changing stressconditions. Data from sections N5, N6, and N7 containing this material were not included in
the calibration data set given low confidence in these backcalculated data due to large
amounts of pavement surface distress in those sections. Equation 6-9 showed a reasonable
model fit (R2 slightly below 0.7) to the calibration data set. This model showed a reasonable
ability to predict the backcalculated moduli from a different data set during the model
validation process. However, this model exhibited similar behavior to the universal model
used for the limerock material in that it tended to over-predict lower modulus values (below
3,000 psi) and under-predict higher modulus values (above 5,000 psi). Despite this, the vast
majority of the data set yielded a reasonable predicted resilient modulus for this material.
Again, a larger data set for model calibration and validation might generate a more robust
model.
2.7613
3.0449
Mr
10465 .42 *14.7 *
14.7
(6-9)
14.7
The base layers that contained the Track soil material for the purposes of backcalculation did
not exhibit strong stress-sensitivity with respect to resilient modulus. As such, no stresssensitivity model could be calibrated to predict the resilient modulus of these materials in the
field. For the Track soil base in sections N8 and N9, the average backcalculated modulus
was 3,942 psi with a standard deviation of 1,109 psi. For the combined Type 5 base and
Track fill used as the composite base layer in section N10, the average backcalculated
modulus was 4,022 psi with a standard deviation of 1,745 psi. For the combined granite base
and Track fill in section S11, the average backcalculated modulus was 11,835 psi with a
standard deviation of 3,384 psi.
The deep Track subgrade behaved similarly to the Track soil material tested in the laboratory
and in the field as a base layer in that it did not exhibit any tangible stress-sensitivity. The
96
average backcalculated modulus for the Track soil material in sections N1 through N4, N10,
and S11 was 34,755 psi with a standard deviation of 7,525 psi. This value compared well to
the average Track subgrade modulus of 32,000 generated via FWD testing during the 2003
research cycle at the Test Track (34).
In sections N8 and N9, the combined Seale and Track subgrade material exhibited tangible
stress-sensitivity in the field. The universal stress-sensitivity model shown in Equation 6-10
was calibrated to predict the moduli of this layer in the field. This model was validated with
another data set and offers a reasonable estimate of backcalculated moduli for this composite
subgrade layer. This stress-sensitivity is likely due to this layer containing approximately 40
inches of the Seale material, which was shown to be very stress-sensitive in laboratory
testing.
1.2255
0.7188
Mr
514.96 *14.7 *
14.7
(6-10)
14.7
Comparison of Laboratory and Field Results

For this project, the five unbound materials were characterized in the laboratory and in the
field under various stress-states and load levels. The result of the laboratory resilient
modulus testing was a constitutive relationship relating resilient modulus to stress-state for
each of the unbound materials. The results of the field resilient modulus testing with the
FWD were constitutive equations for stress-sensitive materials and average modulus values
for non stress-sensitive materials. To compare the laboratory and field-determined resilient
moduli to each other, they must be compared at equivalent stress-states.
First, the stress-states (bulk, deviatoric, octahedral shear) tested in the field at the four FWD
drop heights (6k, 9k, 12k, and 16k) were averaged to determine a representative field stress
for that particular loading. These representative stresses were entered into both the
laboratory-calibrated and field-calibrated constitutive equations to draw comparisons
between the two values for a particular material type. For this comparison, the universal
stress-sensitivity models were used due to their superior R2 values generated during field
calibration. Additionally, these values were plotted against the average and standard
deviation of the backcalculated moduli from the various load levels. This was done to show
the range of backcalculated moduli that were generated due to spatial and construction
variability in the field. For this analysis, dates of FWD testing were analyzed separately.
This was done to eliminate the effects of stress-sensitivity in the modulus variability at the
different load levels, since different pavement temperatures will change the modulus of the
asphalt and consequently alter the stress-states that are experienced by the unbound layers.
For this analysis, comparisons of the results of the laboratory and field testing were only
performed for the materials in which a direct comparison could be made. Several sections
required combining materials for the purposes of generating a viable backcalculation crosssection. As a result, no direct comparison could be made between the laboratory and field-
97
determined resilient moduli for those combined layers. This was true for the combined Seale
and Track subgrade layer in sections N8 and N9, the combined Type 5 and Track soil base
layer for section N10, and the combined granite and Track soil base layer in section S11.
Resilient Modulus (psi) .
Figure 6.9 shows the laboratory versus field-determined moduli for the FWD testing date on
11/27/06 used to generate the backcalculated data for the limerock base material. Figure 6.9
shows that laboratory and field measured data seem to agree well at the lower levels for the
deflection testing on 11/27/06. The laboratory predicted resilient moduli fall within one
standard deviation of the average backcalculated moduli at the 6 kip and 9 kip FWD load
levels. This also shows that the laboratory and field-measured moduli exhibit different
behavior with respect to stress-sensitivity. The laboratory constitutive equation suggests the
material to be stress-hardening while the field-calibrated constitutive equation and
backcalculated moduli suggest the material is stress-softening. This is opposite of what was
expected, since studies have shown unbound materials usually show the same stresssensitivity behavior in the lab and in the field but with very different resilient moduli. This
plot shows reasonable agreement between the field-calibrated constitutive equation and
average backcalculated moduli at the different load levels, with the predicted moduli from
the field-calibrated equation falling within one standard deviation of the average
backcalculated moduli at the various load levels. These results are shown as an example of
the process used to compare laboratory and field data for the base layer materials. Complete
comparison results for all base layer materials are documented in Taylor (27).
Laboratory versus Field Resilient Modulus at Varying FWD Load
Levels - Limerock Base (11/27/06)
40000
Lab Equation
1
35000
Backcalculated Moduli
Field Equation
30000
25000
20000
15000
10000
5000
0
0
3,000
6,000
9,000
12,000
15,000
18,000
FWD Load Level (lb)
FIGURE 6.9 Laboratory versus Field Resilient Moduli Comparison (Limerock Base
Material, FWD Testing on 11/27/06) (27).
The Track soil material was utilized as the deep subgrade material throughout the structural
study at the Test Track. This material was deemed to be non stress-sensitive through both
laboratory and field testing. The average laboratory modulus for this material was calculated
as 28,335 psi with a standard deviation of 6,650 psi. In the field, the average modulus was
98
calculated as 34,755 psi with a standard deviation of 7,525 psi. The average field modulus
was taken from the FWD data calculated in sections N1 through N4, N10, and S11. The data
from sections N5, N6, and N7 were excluded from this data given the larger amounts of
pavement distress in those sections leading to lower confidence in the accuracy of the
deflection data. Sections N8 and N9 were excluded from this data set because the deep
subgrade contained both the Seale and Track soil materials in this section.
A one-way ANOVA test between the two data sets showed that there was a statistical
difference between the means of the two data sets (F-statistic = 58.03). Comparing the
averages of the means yielded a ratio of field moduli to lab moduli of 1.23. This ratio shows
good general agreement between the laboratory and field-measured behavior of the Track
soil material when used as a subgrade. This was expected, since literature indicated that
agreement between lab and field data tends to improve when unbound materials deeper in the
pavement structure are compared (27).
Summary of Findings from Laboratory and Field Modulus Comparisons
For the unbound granular base materials for which comparisons could be drawn (the
limerock base and granite base), poor agreement was seen between the moduli
backcalculated in the field and a modulus predicted from a laboratory-calibrated constitutive
equation over various representative stress-states. For these materials, the laboratory
predicted moduli exhibited stress-hardening behavior and the backcalculated moduli
exhibited stress-softening behavior. This reversal in stress-sensitivity between the two data
sets was opposite of what was seen in literature. Potential reasons for this disagreement
include: laboratory samples that were non-representative of the in-situ conditions
(compaction, density, moisture content, etc.), inherent disagreement between the nature of
the two test methods, and the existence of a very stiff subgrade at the Test Track supporting
and altering the stress-sensitivity behavior of these materials in the field.
Despite the opposing trends, generally good agreement was seen between the lab and field
data at the stresses representative of the lower FWD loading level (6 kip) on multiple testing
dates for both the limerock and the granite base. Given the opposite stress-sensitivity, the
laboratory moduli were often larger than the field moduli at the larger FWD load levels (12
kip and 16 kip). Also, the field-calibrated constitutive equation was shown to be a good
predictor of the unbound material moduli for these two materials at the Test Track.
For the Track soil material, poor agreement was seen between the moduli of the field-tested
base material and the laboratory tested material. The average laboratory modulus was
approximately 7.2 times larger than the average backcalculated base modulus for this
material. This was expected since the Track soil material for this layer was sandwiched
between two much larger layers for the purposes of backcalculation and was supported by the
softer Seale subgrade material. The Track subgrade compared well with the laboratory
values, however. The average backcalculated Track subgrade modulus was only 23 percent
larger than the average laboratory Track soil modulus. This was expected since literature
indicates that agreement between laboratory and field-measured resilient moduli show better
agreement in the deeper pavement layers (47; 48).
99
Based on the results of this comparison, it is recommended that the laboratory constitutive
equation be utilized to characterize the unbound materials at the Test Track for Level 1
MEPDG design. The poor agreement between the laboratory and field resilient moduli is not
necessarily an indicator of low quality lab or field data, but of inherent disagreement between
the testing method and conditions. The field-generated resilient moduli and stress-sensitivity
models were shown to accurately characterize the unbound materials in-situ. However, the
backcalculation data and field-generated stress-sensitivity models are site-specific to the Test
Track. Additional testing at other locations containing the various unbound materials is
advised to further validate the field-calibrated stress-sensitivity models.
Recommendations
This research highlighted the effective use of triaxial testing and FWD testing to give a
mechanistic characterization of the different unbound pavement materials utilized at the Test
Track. The laboratory testing gives a quality representation of the stress-sensitivity of the
resilient moduli for each of these materials. Characterization of the materials in this manner
is sufficient for obtaining a good representation of the unbound material behavior for Level 1
mechanistic pavement design using the new MEPDG. A quality FWD testing program that
encompasses the seasonal and spatial variability of the pavement and tests the pavement at
multiple critical load levels can also be used to develop a constitutive relationship for the insitu unbound material moduli. The equations calibrated for this study are site-specific to the
Test Track, and should be validated through comparison of moduli to additional pavement
structures that contain the given unbound materials, but have varying structural compositions
and thicknesses.
The results of this study showed the following:
Either the MEPDG model or the universal model provide the best fit to laboratory
resilient modulus data, and the universal model provided the best model fit to
backcalculated resilient moduli. Therefore, the multi-variable constitutive models are
recommended over the single-variable models given the results of this study.
It is recommended that the laboratory-calibrated multi-variable stress-sensitivity
equations provide the best means of characterizing the unbound materials used at the
Test Track for Level 1 MEPDG design. However, the FWD testing program at the
Test Track provides a good quantification of the seasonal and spatial variability
inherent to these materials in the field and should be used to account for field
variability in the pavement design process.
100
E* LABORATORY AND MODEL COMPARISONS
There are three levels of design that engineers can choose from when using the MEPDG for
flexible pavement design (49). One significant difference in the three levels of design is how
the dynamic modulus, E*, of HMA is computed and therefore the degree of complexity
required for material property inputs. Level one is the most complex degree of design,
requiring laboratory test results for both E* and dynamic shear modulus of the binder, Gb*.
However, dynamic modulus laboratory testing is time consuming and expensive. Therefore,
many highway agencies will opt to use levels two and three for most pavement designs since
these levels do not require such testing. The required material property inputs for levels two
and three are simple parameters typically determined for specification requirements. Level
two, the second most complex design, utilizes either the 1-37A or the 1-40D Witczak
Predictive E* equation (at the discretion of the user) as a function of binder information,
gradation information, and other volumetric information. Level three, the least complex of
the three designs also utilizes either the 1-40D or the1-37A Witczak Predictive E* equation,
however laboratory binder test results are not required.
Due to the challenges to perform laboratory dynamic modulus tests, there is a need to
validate the MEPDGs procedure in calculating E* at the two lowest levels of design. To
evaluate the accuracy of these two designs, three E* predictive equations were analyzed and
compared with E* laboratory test results. To do so, E* testing was completed on the
materials comprising sections constructed in the 2006 Test Track structural study.
Objectives and Scope
The objectives of this study were to evaluate the three E* predictive models: the Witczak 137A model, the Witczak 1-40D model, and the Hirsch model. Findings from this study
should help pavement engineers to accurately estimate E*.
To meet the objectives of this study, laboratory testing was completed using AASHTO TP
62-07 (49) as a guideline to determine the dynamic modulus of ten different HMA mixtures
from the 2006 NCAT Test Track Structural Study. These mixtures represented typical
mixtures used on state highways in Alabama, Florida, Missouri and Oklahoma. AASHTO
T315-06 (49) was followed to determine the binder complex shear modulus, Gb*. Also,
binder viscosities were obtained following the ASTM D2983-04a (50) for eight binders.
Models to Determine E*
As part of the NCHRP 1-37A project, a predictive model for E* was developed utilizing
rudimentary mix information including basic binder test results to be included in the MEPDG
(51). An initial model was developed in 1996 by Witczak and Fonseca, however it was
quickly updated, expanding the dataset and re-calibrating the model. This model, updated by
Witczak and Andrei, was implemented for use in the MEPDG levels two and three in 1999
(51). This model will be referred to as the Witczak 1-37A E* predictive equation. In 2005,
another equation was developed by Witczak and others under the NCHRP 1-40D initiative
(52). This equation, referred herein as the Witczak 1-40D E* predictive equation, was
implemented into the MEPDG version 1.0 allowing the user to select between the 1-37A and
101
1-40D equation for use in level two and three design. However, a note to users explains that
the 1-40D equation has not yet been nationally calibrated.
In addition to the Witczak equations, another model recently cited by a number of
researchers is the Hirsch E* predictive model (53), developed in 2003. The Witczak 1-37A,
the Witczak 1-40D and the Hirsch E* predictive models were analyzed in this research.
The information required for each model is tabulated in Table 6.5. As mentioned previously,
the level of complexity of the material property inputs varies among the three models. Two
of the three require Gb* testing, while only one requires the phase angle associated with Gb*.
It is worthy of noting that the while both Witczak models incorporate the gradation of the
mix, the Hirsch model does not. In fact, the Hirsch model relies only on three properties of
the mix.
TABLE 6.5
Parameter
Gradation:
P200 Pass.
P4 retained
P38 retained
P34 retained
VMA
Va
VFA
Vbeff
F
Gb*
b
Material Property Requirements by Model

Witczak (1-37A) Witczak (1-40D) Hirsch
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Witczak 1-37A E* Predictive Equation

The original Witczak model was developed in 1996 by Dr. Witczak and his colleagues using
149 unaged HMA mixtures (51). The binder types utilized in the mixtures were conventional
binders only, which severely limits the use of this model. This equation was revised in 1999,
by expanding the dataset and deriving an equation that more accurately fit a wider range of
gradations, binder stiffness and air voids (53). The revised equation, the Witczak 1-37A E*
predictive equation (51), listed in Equation 6-11, and is a function of gradation of the
aggregate, air voids, effective binder content, viscosity of the binder, and loading frequency:
Vbeff
log E*
1.25 0.029 200 0.0018( 200 ) 2 0.0028 4 0.058Va 0.0822
Vbeff Va
3.872 0.0021 4 0.004 38 0.000017 ( 38 ) 2

1 e ( 0.603313 0.313351 log( f ) 0.393532 log( ))
0.0055
34
(6-11)
102
where: E* = dynamic modulus of mix, 105 psi
= viscosity of binder, 106 poise
f = loading frequency, Hz
200 = % passing #200 sieve
4 = cumulative % retained on #4 sieve
38 = cumulative % retained on 3/8 in. sieve
Va = air voids, % by volume
Vbeff = effective binder content
Gb *
4.8628
1
10 sin b
where: Gb* = dynamic shear modulus of binder, psi
b = phase angle of Gb*, degrees
(6-12)
Equation 6-11 is currently used for a level two design in the MEPDG (49). It is also
employed for the level three design with the use of typical viscosity values of the binder
based on the binder grading (49). The viscosity can be determined by conventional binder
tests or from a dynamic shear modulus (Gb*) test. When using the Gb* to determine
viscosity, Gb* and its associated phase angle must be measured at a variety of temperatures
for a loading frequency of 1.59 Hz or 10 rad/sec, from which Equation 6-12 is employed to
calculate viscosity (49). The loading frequency for Gb* testing should not be confused with
the loading frequencies required for the 1-37A model, as the loading frequencies selected for
the model are typical of an E* laboratory test (0.1, 1, 10, 25Hz, etc.). Viscosity can also be
estimated from conventional binder tests. However some methods require a conversion to
achieve a viscosity in Poises.
Witczak 1-40D E* Predictive Equation
The Witczak 1-37A E* predictive model was revised in 2006, again expanding the dataset
used to develop the original model, using a total of 346 HMA mixes to calibrate the model
(51). One important difference to note between the 1-37A E* predictive model and the 1-40D
E* predictive model is the replacement of the viscosity and loading frequency parameters
with the complex shear modulus of the binder, Gb*, and its associated phase angle, b, as
shown in Equation 6-13:
103
6.65 0.032
log E*
0.349 0.754 Gb *
0.0052
0.0001(
0.08Va 1.06
2.56 0.03Va
0.71
Vbeff
Va
1 e
Vbeff
0.012
38
0.0027 (
200
0.006
)2
0.011
0.00014 (
38
4
2
Vbeff
Va
0.0001(
( 0.7814 0.5785 log Gb * 0.8834 log
38
200
38
Vbeff
)2
0.01
34
b)
(6-13)
where: E* = dynamic modulus of mix, psi
Gb*= complex shear modulus of binder, psi
200 = % passing #200 sieve
4 = cumulative % retained on #4 sieve
Va = air voids, % by volume
Vbeff = effective binder content
b = phase angle of binder associated with Gb*, degrees
This equation, referred herein as the Witczak 1-40D E* predictive equation, was
implemented in the MEPDG for use in level two and three design. As mentioned previously,
the user as the option of selecting the 1-37A or the 1-40D model for design. At a level three
design typical Gb* and b values are estimated within the program depending on the selected
grade of the binder. Although the loading frequency parameter has been omitted in the new
model, the time and temperature dependency of dynamic modulus is characterized by Gb*
and b.
Hirsch E* Predictive Model
The Hirsch E* predictive equation is similar to the Witczak equations in that it also utilizes
volumetric information and Gb* laboratory test results. However, this model requires only
two volumetric properties: voids in mineral aggregate (VMA) and VFA. The Hirsch E*
model was developed based on the law of mixtures, also referred to as the Hirsch model, for
composite materials (54). The Hirsch E* predictive equation for asphalt mixtures, as
developed by Christensen and his colleagues is:
E * mix
Pc 4,200,000 1
VMA
100
3G *b
VFA VMA
10,000
1 PC
VMA/ 100
4,200,000
VMA
3VFA G * b
104
where:
20
Pc
650
VFA 3 G * b
0.58
VMA
VFA 3 G * b
0.58
VMA
(6-14)
where: E*mix = dynamic modulus, psi

VMA = voids in mineral aggregate, %
VFA = voids in aggregate filled with mastic, %
VFA = 100*(VMA-Va)/VMA
Va = air voids, %
G*b = complex shear modulus of binder, psi
Just as in the 1-40D model, the time and temperature dependency of E* is characterized in
the Hirsch E* predictive model by Gb*. According to Christensen, the Gb* should be at the
same temperature and loading time selected for the mixture modulus, and in consistent units
(54). For this evaluation, it was assumed that the frequency-time relationship is consistent
for both Gb* and E* tests such that a selected frequency applies the load for the same time in
either test. Therefore, the loading frequencies and temperatures selected associated with Gb*
were the same as those from the measured E* values, which enabled a direct comparison of
predicted to measured E* values.
Testing Protocol
To complete the dynamic modulus testing for each unique mix listed, the testing protocol
outlined by AASHTO TP 62-07 was followed (49). An AMPT, shown in Figure 6.11, was
utilized to apply haversine compressive loading for a range of frequencies and temperatures.
For this investigation E* results at three temperatures, 40, 70, and 100F and seven
frequencies, 0.5, 1, 2, 5, 10, 20, and 25 Hz were obtained.
105
FIGURE 6.11 The AMPT Machine and Close Up of Specimen.

Viscosity was obtained for various binders using the Brookfield Viscometer following
ASTM D2983-04a (50). The binders were aged in a Rolling Thin-Film Oven (RTFO) and
viscosities were obtained at two temperatures, 135 and 165C. After the test temperatures
were converted to Rankine, Equation 6-15 was utilized to extrapolate viscosity for the
temperatures (40, 70, and 100F) at which E* testing was completed.
log log
A VTS log TR
(6-15)
where: = viscosity of binder, centipoise (cP)
A, VTS = regression parameters
TR = temperature, Rankine
Complex shear modulus, Gb*, testing, in accordance with AASHTO T 315-06 (49), was also
conducted on RTFO aged binders using a Dynamic Shear Rheometer (DSR). Although
testing was conducted at four temperatures, 4, 21, 37.8, and 54.4C (40, 70, 100, and 130F),
results at 40F were inconsistent and unreliable and thus excluded from the analysis. A
frequency sweep, in which thirteen frequencies (0.1, 0.2, 0.3, 0.4, 0.6, 1.0, 1.6, 2.5, 4.0, 6.3,
10.0, 15.9, and 25 Hz) were applied, was completed at each of the four temperatures.
Mixtures Tested
E* laboratory testing was completed for the mixes in each section constructed in 2006 as part
of the 2006 Test Track structural study. The mixes investigated are described in Table 6.6,
listed by the section and layer in which it was placed. Four different mix types were
incorporated in the investigation including Superpave mixes (super), SMA mixes, rich
bottom layers (RBL) and a typical mix design used by the Oklahoma E* laboratory testing
was completed for the mixes in each section constructed in 2006 as part of the 2006 Test
Track structural study. The mixes investigated are described in Table 6.6, listed by the
section and layer in which it was placed. Four different mix types were incorporated in the
106
investigation including Superpave mixes (Super), stone matrix asphalt mixes (SMA), rich
bottom layers (RBL) and a typical mix design used by the Oklahoma DOT, designated by
S3. Each mixture is described by a unique mix number and unique binder number. The
gradation information acquired from as-built records, as well as the NMAS and gradation
type (coarse/fine) for each mix number is listed in Table 3. As shown in Table 6-6, a total of
ten unique mixes were evaluated, however, some unique mixes were used in multiple layers.
Associated with a unique binder number is the PG grade, viscosity, and G* test results
(including phase angle values, b). An example is provided in Table 6.7. The other input
values are documented elsewhere (55). A total of nine binders were used in the HMA layers
investigated, and similar to the unique mix number, some unique binders were used in
multiple layers.
Viscosity values were not obtained for unique binder #4, which was used only in mix #7.
From these data, the dynamic moduli of the mixes, except mix #7, were predicted for the 137A E* predictive model at three temperatures and seven loading frequencies. Gb* and b
values were obtained at multiple temperatures and frequencies. However, to compare
laboratory measured E* values with E* predictions from either the Hirsch or 1-40D E*
predictive models, the loading frequencies and test temperatures for Gb* must equal to those
used in the E* laboratory test. As a result, the 1-40D and Hirsch E* predictive models were
evaluated at two temperatures, 70 and 100F, and three frequencies, 1, 10, and 25 Hz.
Results and Discussion
In following Equation 6-11 to evaluate the Witczak 1-37A E* model, 651 data points were
produced. The values were plotted against the laboratory measured E* values (Figure 6.12).
For the Witczak 1-40D and Hirsch E* models, 180 data points were evaluated for each and
also plotted against measured E* values. Ideally, all points would fall along the line of
equality, but this clearly is not the case. Upon visual inspection, it can be concluded that
overall the 1-40D E* model consistently lies above this line, indicating a consistent
overprediction of the measured values. While there is a large amount of scatter in the data,
the 1-37A E* model generally follows the line of equality, with data points both above and
below this line. Similarly, the Hirsch E* model also generally follows the line of equality,
although there is much less scatter in these data.
Fitting a linear regression equation indicates how precisely the modeled values are linearly
related to the measured values. Some researchers (51; 53) use the coefficient of
determination, R2, from as an indication of the quality of the relationship. Inset in Figure 2
are the sample sizes, linear regression equation and associated coefficient of determination in
log scale for each model evaluated. All three models produced predictions that were fairly
correlated to the measured values by a linear relationship. Overall, these values may be
slightly skewed due to the gross overprediction of one mixture. The measured values for this
mix are concerning as they are relatively low, on the order of 10,000+ psi, which could be an
indication of an error in that particular test. It could also be an indication of bias in all three
models. The Hirsch E* model produced the highest R2 value, 0.707, while the Witczak 140D produced the lowest, 0.573 in log scale. This is contrary to previous findings in which
the 1-40D model was found to be an improvement over both the Hirsch and 1-37A model
(51; 53).
107
TABLE 6.6 HMA Mixes by Section and Layer
Section
N1
N1
N1
N2
N2
N2
N8
N8
N8
N8
N9
N9
N9
N9
N9
N10
N10
S11
S11
S11
Layer
1
2
3
1
2
3
1
2
3
4
1
2
3
4
5
1
3
1
3
4
Sponsor Mix Type

FL
Super
FL
Super
FL
Super
FL
Super
FL
Super
FL
Super
OK
SMA
OK
S3
OK
S3
OK
RBL
OK
SMA
OK
S3
OK
S3
OK
S3
OK
RBL
MO
Super
MO
Super
AL
Super
AL
Super
AL
Super
Binder
67-22
67-22
67-22
76-22
76-22
67-22
76-28
76-28
64-22
64-22
76-28
76-28
64-22
64-22
64-22
70-22
64-22
76-22
67-22
67-22
Unique
Binder #
1
1
14
2A
2A
14
3
3
4A
4
3
3
4A
4A
4
5
6
13
14
14
Unique
Mix #
1
1
2
3
3
2
4
5
6
7
4
5
6
6
7
8A
9
28A
2
2
TABLE 6.7 Gradation Information by Mix #

Unique
Mix #
1
2
3
4
5
6
7
9
8A
28A
Unique
Binder #
1
14
2A
3
3
4A
4
6
5
13
200 %
passing
8.79
5.48
8.10
10.71
6.90
6.90
10.51
6.28
5.40
8.59
4%
retained
39.61
45.82
38.61
68.55
34.80
36.31
39.93
51.79
47.71
13.98
8%
retained
16.44
25.19
16.62
28.62
19.53
21.13
13.68
25.60
16.83
0.10
34 %
retained
0.00
4.47
0.00
0.00
4.71
5.48
0.00
1.81
0.79
0.00
VMA
(Avg)
14.17
15.41
14.22
15.57
10.46
10.69
12.44
14.17
16.92
18.05
NMAS
(mm)
12.5
19
12.5
12.5
19
19
12.5
19
12.5
9.5
Graded
Fine
Fine
Fine
Coarse
Fine
Fine
Coarse
Fine
Coarse
Fine
However, the coefficient of determination is a measure of how precisely the linear regression
equation matches the data. For the model to be accurate, the linear regression equation should
be as close to y = x as possible, therefore the slope and intercepts of the regression equations
should also be investigated. Table 6.8 lists the linear regression coefficients and goodness of
fit in arithmetic scale for each of the models.
108
Evaluation of E* Models on Structural Test Track Sections

7.0
1-37A
1-40D
Hirsch
Predicted Log E* (psi)
6.5
6.0
5.5
Witczak 1-37A:
n = 651
y = 0.8142x+0.9639
R2 =0.6846
Witczak 1-40D:
n = 180
y = 0.4827x+3.3128
R2 =0.5728
Hirsch:
n = 180
y = 0.612x+2.1791
R2 = 0.7071
5.0
4.5
4.0
3.5
3.5
4.0
4.5
5.0
5.5
6.0
6.5
7.0
Measured Log E* (psi)
FIGURE 6.12 Comparison of Predicted E* Values to Measured E* Values by Model.

TABLE 6.8 Linear Regression Coefficients for Each Model
Model
Witczak 1-37A
Hirsch
Witczak 1-40D
Slope
1.1222
0.6354
1.2748
Intercept (psi)
56,263
144,315
515,267
R
0.5945
0.766
0.5996
According to these lines of best fit, the Witczak 1-37A has the smallest difference in slopes
from the line of equality, deviating above the line by only 12.22%. Additionally, the yintercept is the smallest of the three models. However, indicated by the R2 and Figure 6-12,
the 1-37A E* model does not consistently predict the measured values accurately. The Hirsch
model is more consistent than the Witczak 1-40D model, however, the slope for the Hirsch
model deviates from unity the most. In looking at Figure 6-12, the data for both the Hirsch
and 1-40D models appear to flatten out at intermediate measured moduli values (700,0001,400,000 psi). Given that both of these models follow this trend and utilize Gb*, such
inconsistencies could be related to the use of Gb* values, whether it be inherent to the model
or Gb* test results. This range of measured E* is consistent with values predicted from Gb*
test results at 70F. This trend is contrary to the time-dependency of HMA, such that
increases in the frequency should result in higher E* values. When the data are plotted in
arithmetic scale, the flattening out of the data at higher moduli values is more predominant in
the Hirsch model than the 1-40D model, which may explain the large deviation in slope from
the line of equality. In general, it appears that the Hirsch E* model predicts the measured
values most accurately and precisely for low dynamic moduli (100,000-500,000 psi).
Consistent with previous research (56), the 1-40D E* model grossly over predicts dynamic
109
modulus at low values or high temperatures and/or low frequencies. The Witczak 1-40D E*
model, is only slightly more precise than the 1-37A E* model.
Summary
Three E* predictive models were evaluated for asphalt mixtures from the southeastern U.S.
The dynamic moduli of ten different mixtures from the 2006 Test Track were measured in
the laboratory and compared with moduli predicted by the 1-37A, 1-40-D, and Hirsch E*
predictive models. Following the Witczak 1-37A E* model, E* was estimated for three
temperatures (40, 70, and 100F) and seven frequencies (0.1, 1, 2, 5, 10, 20, and 25 Hz). E*
was also estimated using the Witczak 1-40D and Hirsch E* predictive models for two
temperatures (70, and 100F) and three frequencies (1, 10, 25 Hz). The findings from these
comparisons are summarized below:
The Witczak 1-37A E* model was found to be unreliable with a large amount of
scatter.
The Hirsch E* model was found to be the most precise model, with the highest
coefficient of determination, 0.707 in log scale and 0.766 in arithmetic scale.
The Witczak 1-40D E* model consistently overpredicted E* particularly for low
dynamic moduli.
The Hirsch E* model is most accurate at low dynamic moduli values (100,000700,000 psi).
At an intermediate test temperature (70F), both the Witczak 1-40D E* and Hirsch E*
models flatten out, thus misrepresenting the time-dependency of E*.
In the current version of the MEPDG, Witczak E* predictive equations 1-40D and 1-37A, are
employed to determine dynamic modulus given volumetric, gradation and binder properties
of a mixtures. In applying these models to ten mixtures included in the 2006 Test Track
structural study, neither model consistently estimated laboratory dynamic moduli values to a
high degree of accuracy. Because both the 1-37A and 1-40D models were found to be
unreliable and the 1-40D model largely overpredicts dynamic modulus, it is recommended
that the Hirsch E* model be used. It should be used with caution however, as discrepancies at
lower temperatures and/or higher frequencies were reported.
LABORATORY FATIGUE THRESHOLDS AND FIELD-MEASURED STRAINS
Laboratory work has been conducted to validate fatigue thresholds for perpetual pavements;
however, little has been published to relate these laboratory values to what is experienced in
the field. Are there pavement structures that have shown superior performance while
undergoing strains greater than the proposed laboratory-based limits? Have field pavements
exhibited pavement responses smaller than the proposed strain limit and failed prematurely?
Currently, field thresholds for fatigue cracking are only conservative estimates based upon
laboratory work. It is possible the field thresholds for fatigue cracking will be higher than
those seen in laboratory work. If this is the case, engineers are overdesigning their
pavements. This analysis compares the strain distributions developed from the previously
mentioned methodologies (Chapter 5) to laboratory fatigue endurance limits tested under the
National Cooperative Highway Research Program (NCHRP) Project 9-38.
110
Laboratory Testing
As part of NCHRP Project 9-38, laboratory fatigue tests were conducted on the base mixes
from the structural sections at the 2003 NCAT Test Track. Beam fatigue specimens were
fabricated using the average gradation and asphalt contents of the 19.0 mm mixes. As
previously seen in Chapter 4, two different mixes were used as the base HMA layers in the
first seven sections of the 2003 experiment. Therefore, only two mixes were tested to
characterize all seven test sections. The tests used samples compacted to 7% air voids. The
detailed methodologies used for the testing are documented elsewhere (57).The fatigue
endurance limits and their 95% lower bound confidence limits are presented in Table 6.9.
These results were extrapolated from laboratory testing using a Three-Stage Weibull
equation (57).
TABLE 6.9 Fatigue Endurance Limits for 2003 Test Sections (29)
Average Extrapolated Beam
95% Confidence Lower
Section
Fatigue,
Bound,
N1
220
146
N2
172
151
N3
172
151
N4
220
146
N5
220
146
N6
172
151
N7
172
151
Laboratory Testing and Field Data Comparisons
The purpose of this analysis was to determine if a relationship existed between the laboratory
fatigue endurance limits determined under NCHRP 9-38 and the cumulative strain
distributions developed in the previous chapters. This was accomplished using three phases
of data comparisons.
The first phase was conducted by comparing the endurance limit to a strain calculated using
the strain-temperature equations developed in Chapters 5. The comparison temperature was
the prescribed testing temperature for AASHTO T321.
The second phase compared the magnitude of the laboratory fatigue threshold data to
measured field strain distributions. The 95th percentile confidence interval lower bound
strain was graphically inserted onto a cumulative distribution plot to determine where this
value fell on each sections cumulative strain distribution, and that value was compared to
the sections fatigue performance.
The final analysis comparison of the laboratory fatigue thresholds and the developed strain
distributions was conducted by comparing the 95th percentile confidence interval lower
bound to the entire strain distribution. A fatigue ratio, between the measured strain values
and laboratory-established threshold, was developed to make this comparison.
111
The 95% confidence interval lower bound was chosen to be the basis of comparison between
the laboratory fatigue and field strain data. 220 , the predicted beam fatigue value for the
PG 76-22 binder mix, was more than double the previously listed fatigue limits in mixes.
This value might be due to a high reading or an erroneous test; therefore, it was felt that
using the 95% confidence lower bound was more appropriate for comparisons. The 95%
confidence interval lower bound also brought the inclusion of mix variability to the analysis;
therefore, it proved to be a more conservative estimate of the fatigue threshold.
Phase One Comparison of Fatigue Threshold to Modeled Strain Levels
The laboratory beam fatigue tests were conducted at 20C (68F). Comparative inquiries
were conducted to calculate the estimated strains at the test temperature for both the 2003
and 2006 Test Tracks using the strain-temperature relationships previously developed and
discussed in Chapter 5. These strains are presented in Table 6.10.
When analyzing the 2003 Test Track estimated strain relationships, one would see that at
20C, five of the six sections average box trailer produced a microstrain magnitude below its
lower confidence limit fatigue threshold. However, when the heavier triple trailers were
used, six of the seven test sections were above the 95th percentile threshold strain. While
two of the 2003 test sections did not fail in fatigue, one of these two sections had a strain
magnitude larger than its fatigue threshold limit at the associated test temperature.
While fatigue testing has not been conducted on the base HMA layers of the 2006 Test Track
at this time, if one were to assume similar lower bounds for the fatigue thresholds (i.e. 150
microstrain) for these sections, a similar trend would be seen. Only two sections (N10 and
S11) experienced higher strains than the lower bound fatigue limit under the steer axle. The
three sections that did not crack were at least 60 microstrain below this value for their steer
axles. However, when analyzing the 40 kip tandem axle at the test temperature, the only
sections still below the estimated lower bound fatigue endurance limit were the three sections
that did not crack.
A final look at the single axles for the 2006 Test Track shows seven of the eight sections
measuring strains higher than the lower bound fatigue endurance limit. The only section
measuring strain amplitudes below this value is section N9 which was 14 inches thick. To
understand this phenomenon, one would only have to look back to the strain-temperature
relationships developed for this section. Both the k1 and k2 values were small compared to
the other sections. This showed that section N9, unlike many of the others, did not
experience as much strain variation due to temperature.
112
TABLE 6.10 Strains at 20 C Estimated from Field Measurements (29)
Strain at 20C
Section
(Year)
N1 (2003)
N2 (2003)
N3 (2003)
N4 (2003)
N5 (2003)
N6 (2003)
N7 (2003)
N1 (2006)
N2 (2006)
N3 (2006)
N4 (2006)
N8 (2006)
N9 (2006)
N10 (2006)
Box,
NA
142
103
87
99
165
111
Steer,
137
148
90
81
135
46
186
Triple,
363
231
153
140
172
210
159
Tandem,
195
203
147
137
173
69
233
Single,
221
228
175
163
185
82
279
S11 (2006)
181
265
294
Phase One - Analysis Results

No clear relationship could be developed between the 95th percentile confidence interval
lower bound, the computed strain at 20 C, and pavement performance. There was a case
where a test section (N3 2003) did not exhibit fatigue cracking; however, its average strain
for the triple trailer at laboratory testing temperature was above the designated fatigue
threshold. Similar results were seen for the tandem and single axles of N1 and N2 in 2006.
Further investigations needed to be conducted to develop the link between laboratory fatigue
thresholds, measured field strains, and pavement performance.
113
Phase Two Location of Fatigue Thresholds on Cumulative Distribution Plots

Another method for comparing the laboratory fatigue thresholds to the previously developed
strain data was to superimpose the 95th percentile confidence interval lower bound strain onto
each sections cumulative strain distribution. The cumulative distributions and fatigue
threshold for the PG 67-22 mix are shown in Figure 6.13 while Figure 6.14 shows the data
for the PG 76-22 mixes.
When comparing the 2003 strain distributions with their lower bound fatigue limit, it was
difficult to find a relationship between the laboratory and field data using this methodology.
As can be seen, there is no clear correlation between the lab fatigue threshold, where this
threshold occurs on the strain distribution, and the sections performance.
For example, Section N4 in 2003 had 40% of its strains below the given laboratory fatigue
threshold, and it performed well in the field. Conversely, section N2 had 45% of its strains
below the laboratory fatigue threshold, but it failed quickly due to fatigue cracking.
Fatigue Threshold.
100%
90%
80%
Percentile
70%
60%
50%
40%
30%
N2 2003 (Cracking)
N6 2003 (Cracking)
N3 2003 (No Cracking)
N7 2003 (Cracking)
20%
10%
0%
0
200
400
600
Microstrain
FIGURE 6.13 Endurance Limits for PG 67-22 Mix (29).
114
800
1000
Fatigue Threshold.
100%
90%
80%
Percentile
70%
60%
50%
40%
30%
N1 2003 (Cracking)
20%
N4 2003 (No Cracking)

N5 2003 (Cracking)
10%
0%
0
200
400
600
800
1000
1200
1400
Microstrain
FIGURE 6.14 Endurance Limits for PG 76-22 Mix (29).

Both base HMA mixes for the 2003 Test Track had similar fatigue properties (i.e. fatigue
thresholds near 150 ). Since fatigue data had not been developed from the 2006 Test
Tracks base mixes, it was assumed that these mixes had similar fatigue properties for
continuation of the analysis. Figure 6.15 displays these data.
The 2006 data finds similar discontinuities between the field and laboratory using this
methodology. N2 experienced no fatigue cracking; however, it had fewer strains measured
below its fatigue threshold than did section N8 which experienced fatigue distress.
Phase Two Analysis Results
The goal of this analysis was to determine if a relationship existed between the laboratory
fatigue thresholds, the measured strain data, and pavement performance. In order to correlate
these three parameters together for comparison across the analyzed test sections, Table 6.11
was developed.
115
100%
90%
80%
Percentile
70%
N1 (Top-down)
60%
N2 (Top-down)
50%
N3 (No crackng)
N4 (No cracking)
40%
N8 (Cracked)
N9 (No cracking)
30%
N10 (Cracked)
S11 (Cracked)
20%
10%
0%
0
200
400
600
800
Microstrain
1000
1200
1400
FIGURE 6.15 Endurance Limits for 2006 Test Sections (29).

TABLE 6.11 Comparison of Cumulative Strain Distribution and 95% Confidence
Interval Lower Bound (29)
Section Percentile of Fatigue Threshold
Performance
N1 2003
8th
Cracked
th
N2 2003
45
Cracked
th
N3 2003
57
No cracking
N4 2003
40th
No cracking
th
N5 2003
37
Cracked
th
N6 2003
29
Cracked
N7 2003
42nd
Cracked
st
N1 2006
31
Top-down cracking
nd
N2 2006
22
Top-down cracking
N3 2006
33rd
No cracking
th
N4 2006
38
No cracking
N8 2006
31st
Cracked
th
N9 2006
74
No cracking
th
N10 2006
11
Cracked
S11 2006
8th
Cracked
116
As seen in this table, there is no correlation between the percentile where the fatigue
threshold occurred along a sections cumulative distribution and its fatigue performance;
however, a trend was noticed in certain cumulative strain distributions.
If one were to specifically compare N2 and N4 from 2003 again, no correlation between the
locations of the fatigue threshold on their cumulative strain distributions and performance
could be developed. However, above the fatigue threshold, N2s strain distribution begins to
flatten out while N4s strain distribution continues on a gradual slope. Since the lower
strains were similar between the two sections, it seemed the strains above the fatigue
threshold contributed to the deterioration of section N2. This makes sense since one would
expect that higher strains contribute more to pavement damage. Therefore, a third analysis
phase was developed to try to bridge the laboratory and field data using the entire cumulative
distribution rather than just particular points on the distribution.
Phase Three Ratio Comparison
While the two previous phases this research analysis did not find any relationship between
field and laboratory fatigue data, both analysis procedures were limited to one point either on
the cumulative strain distribution or on the strain-temperature relationship curves. The third
phase of comparisons included comparing the entire cumulative strain distribution of each
section to the fatigue threshold by the use of a fatigue ratio (Equation 6-16).
Rn
(6-16)
Where: Rn = ratio at the nth percentile

n = strain at the nth percentile, microstrain
f = fatigue threshold, microstrain
This ratio was calculated for each test section from the 50th to the 99th percentile on a 5
percent increment. These ratios are tabulated in Tables 6.12 and 6.13.
TABLE 6.12 Ratio by Percentile for 2003 Test Sections (29)
Percentile
Cracked
99%
95%
90%
85%
80%
75%
70%
65%
60%
55%
50%
N1
2003
Yes
3.22
2.93
2.67
2.47
2.30
2.22
2.15
2.08
2.02
1.96
1.90
N2
2003
Yes
5.86
4.64
3.88
3.31
2.85
2.53
2.22
1.96
1.66
1.34
1.13
N3
2003
No
2.63
2.34
2.11
1.92
1.79
1.69
1.60
1.50
1.40
1.32
1.23
N4
2003
No
2.48
2.21
2.00
1.82
1.70
1.59
1.48
1.40
1.31
1.23
1.15
N5
2003
Yes
3.56
3.06
2.70
2.43
2.22
2.04
1.88
1.75
1.61
1.47
1.33
117
N6
2003
Yes
4.36
3.73
3.27
2.96
2.69
2.48
2.30
2.14
1.98
1.81
1.65
N7
2003
Yes
4.04
3.37
2.89
2.57
2.30
2.09
1.91
1.75
1.60
1.45
1.30
TABLE 6.13 Ratio by Percentile for 2006 Test Sections (29)
Percentile
Cracked
99%
95%
90%
85%
80%
75%
70%
65%
60%
55%
50%
N1
2006
Yes*
2.36
1.95
1.72
1.57
1.46
1.39
1.33
1.29
1.24
1.21
1.16
N2
2006
Yes*
2.60
2.17
1.94
1.77
1.63
1.53
1.45
1.40
1.35
1.30
1.26
N3
2006
No
2.83
2.45
2.18
1.98
1.85
1.74
1.63
1.53
1.44
1.35
1.27
N4
2006
No
2.54
2.22
1.98
1.80
1.68
1.58
1.48
1.39
1.31
1.23
1.16
N8
2006
Yes
6.21
4.83
3.95
3.29
2.82
2.45
2.18
1.95
1.76
1.56
1.39
N9
2006
No
1.90
1.64
1.42
1.24
1.10
1.01
0.92
0.83
0.75
0.68
0.61
N10
2006
Yes
7.33
5.97
4.96
4.34
3.84
3.45
3.10
2.80
2.54
2.30
2.08
S11
2006
Yes
7.91
6.45
5.57
4.94
4.40
4.00
3.64
3.30
2.99
2.68
2.37
* top-down cracking only

It was seen that a distinct difference was found between the ratios of the sections that failed
and the ratios of those that did not. At the 99th percentile, all the sections exhibiting fatigue
cracking had ratios greater than 3.2, and the sections that performed well were all under 2.85.
This separation between the ratios of cracked and uncracked sections continues until the 55th
percentile.
Phase Three Analysis Results
Based upon these limited data, it is proposed that control points could potentially be set along
strain distributions using the fatigue ratio to help eliminate fatigue cracking for perpetual
pavement designs. These control points (Table 6.14) were based upon the section that
remained in-tact while returning the highest fatigue ratio during the 2003 and 2006 Test
Track analyses (i.e., N3 using the by axle analysis).
TABLE 6.14 Fatigue Control Points for Fatigue Crack Prevention (29)
Percentile
99%
95%
90%
85%
80%
75%
70%
65%
60%
55%
50%
Maximum Fatigue Ratio

2.83
2.45
2.18
1.98
1.85
1.74
1.63
1.53
1.44
1.35
1.27
118
Fatigue Threshold Summary

While it was difficult to determine a relationship between the laboratory fatigue threshold,
field pavement responses, and pavement performance using just one location on either a
strain distribution or strain-temperature model, it was possible to propose a new perpetual
pavement design concept using the entire strain distribution as a basis for comparison. The
following results were established in this research.
A relationship could not be developed between the field-measured strain at the
laboratory testing temperature (20C) and field performance.
The laboratory fatigue threshold did not consistently occur on a pavements
cumulative strain distribution in similar locations for cracked or uncracked
pavements.
While more data are needed from differing mix designs to validate the concept of a
fatigue ratio, data from the 2003 and 2006 Test Tracks support the concept of using
control points set by multiples of the laboratory fatigue 95th percentile confidence
interval lower bound to govern fatigue design.
119
CHAPTER 7 FINDINGS AND STATE DOT IMPLEMENTATIONS
The Test Track operates as a facility where highway owners and the private industry sector
can test new pavement design concepts and technology in a controlled and monitored
environment. For the test track to be beneficial to its sponsors, the research findings need to
be implemented by the sponsoring agencies. The following paragraphs describe how
sponsors have used past Track research findings and implemented the results into their
practices and plan to make use of the current Track research as well.
Mechanistic-Empirical Pavement Design
Through sponsorship of instrumented structural experiment sections, ALDOT, FDOT,
MODOT, ODOT, and FHWA are helping refine the M-E pavement design approach and
facilitate implementation of the MEPDG. Detailed laboratory testing in conjunction with
surface performance measurements are essential in the calibration of the MEPDG using local
materials and methods. Pavement response measurements in these instrumented test sections
provide refinements or validation of the response models in the MEPDG as well as a
calibration mechanism for other M-E design methodologies that output pavement response
(e.g., PerRoad). Each sponsor's individual study provides the necessary connection to local
materials, and data from the entire group of experiments provides an opportunity to test M-E
pavement analysis and design using a broad range of materials and structural buildups.
Agencies can have confidence in the utility of M-E pavement design as a result of the highly
controlled experiment at the Test Track.
Alabama
The Alabama DOT has proactively used the Track in all three completed research cycles to
refine specifications and introduce new methodologies aimed at reducing the life cycle cost
of pavements. When Superpave was introduced, its requirements encouraged coarser-graded
mixes. However, these mixes were susceptible to permeability problems. Multiple
experiments on the 2000 Track proved that fine gradations could perform as well as coarse
gradations in the field. These experiments encouraged Alabama and other agencies to
change their specifications allowing requiring fine-graded mixes to be placed on their
high-volume roadways. In Alabama, upper binder and wearing courses (within 4" {100 mm}
of the surface, not counting OGFC) are no longer allowed to be designed on the course side
of the restricted zone.
Florida
Prior to the 2000 Track research results, the Florida DOT also required mixes to be designed
on the coarse side of the maximum density line (MDL) for all heavy traffic applications.
When fine mixes were found to exhibit comparable performance, FDOT specifications were
revised to allow their use. Now, over 90 percent of mixes in the FDOT QC database are finegraded mixes, which is an indication of the popularity of fine blends within the contracting
community.
The 2003 research cycle was used to validate results from FDOT's heavy vehicle simulator
(HVS). This research found that mixes produced with polymer-modified PG 76 binder rutted
120
at half the rate of the same mix produced with unmodified PG 67 binder. This finding was
significant for several reasons. The validation of results from the HVS research program
made it possible to apply other polymer versus non-polymer findings. The outcome of this
effort was a specification for the use of polymer modified binders in various layer
combinations as a function of traffic in order to optimize the cost of construction.
In 2006, FDOT sponsored research that had two objectives. The first objective was to help
advance understanding of mechanistic pavement designs by building two instrumented
structural sections using a pavement thickness and materials commonly used in Florida. The
other objective was to validate the energy ratio method developed at the University of Florida
to predict mixes prone to top-down cracking. Prior to construction, laboratory testing and
analysis following the energy ration concept indicated that by changing surface mix from a
PG 67 to a polymer modified PG 76 binder would yield a substantial improvement in
cracking resistance. As predicted, it took the application of approximately 50 percent more
ESALs (2.9 million in N2 versus 1.9 million in N1) in order to induce surface cracking
throughout the length of the PG 76 section. In conjunction with the Track study, FDOT also
sponsored a project at the University of Florida to refine the laboratory procedure for their
mix design evaluation process. When the simplified equipment is made available, FDOT
plans to use the validated approach to either approve mixes on a case by case basis or to
establish guidelines that can be used to ensure new mix designs will be resistant to surface
cracking.
Another outcome of the FDOT experiment was that strains measured at the bottom of the two
pavement structures were essentially identical until the surface layer cracked in the
unmodified section. Based on these results, it can be inferred that surface mixes with
modified binders does not affect strain performance at the bottom of the HMA layers.
Georgia
In past research cycles, GDOT sponsored research on the Test Track to compare rutting and
durability performance of more expensive SMA surfaces to dense-graded Superpave mixes.
Although the dense-graded mix exhibited excellent rutting performance that was statistically
equivalent to the SMA mix, the SMA surface proved to be more resistant to raveling and
general degradation associated with surface aging. As a result of these findings, GDOT
implemented a policy whereby lower volume roadway surfaces that would otherwise have
been resurfaced using a drainable surface mix (to optimize safety) over an SMA (to optimize
performance) were instead resurfaced using a drainable surface over a less costly densegraded Superpave mix (to reduce the cost of construction). This change in practice was
expected to offer the same rutting performance as SMA, while at the same time eliminating
the durability shortcoming of the dense mix by protecting it with a permeable surface. A
significant cost savings resulted from the substitution of dense mix for SMA without
compromising performance.
In order to ensure a stable aggregate structure in permeable surface mixes, GDOT specifies a
5:1 flat and elongated limit of 10 percent on the coarse aggregate fraction of the mix. GDOT
sponsored test sections on the 2006 Test Track to investigate the possibility of changing the
flat and elongated specification to 3:1 in order to improve drainage properties. It was
121
observed that the porous surface mix containing an aggregate having more flat and elongated
particles actually improved the drainability of the pavement surface, an effect that was
measured both in the laboratory and on the surface of the Track. The performance of the
comparative drainable surfaces was equal in every other way.
Another GDOT test section was also built with a double-layer porous surface course. The
lower layer contained the coarse aggregate with more flat and elongated particles, and the
upper layer used a smaller NMAS open-graded mix. The two porous mixes were laid
simultaneously using a European twin layer paver. In many urban areas, such as Atlanta,
Georgia, federal policy requires the installation of sound barriers in order to minimize the
effect of traffic noise on nearby communities. It has been observed in Europe that smaller
NMAS porous mixes placed simultaneously on larger NMAS porous mixes creates an
acoustic condition that is favorable to noise reduction. Additionally, the smaller NMAS
surface mix may provide a filtering mechanism that would promote the long-term
drainability of the larger NMAS mix. Since tack coat and construction traffic could hinder
the permeability of the porous mixes, the European paver was developed to place both layers
in a single process without the use of tack. GDOT's twin layer section on the NCAT
Pavement Test Track proved to be the most drainable and quietest surface from start to finish
over the entire 2-year (10 million ESAL) research cycle. Ideally, this section will remain in
place for the 2009 research cycle in order to document the long term (drainability and noisereducing) performance of twin layer pavements.
Indiana
INDOT has reported unexpectedly high variability for air voids of plant produced HMA.
Some QC data has shown air voids 4% from the job mix formula compared to the specified
1%. Indiana developed an experimental program which would allow it to quantify the
performance of low air void pavements in the field in order to validate new acceptance
criteria for pavements. Pavements were built with air voids ranging from 0 to 4% air voids
and rutting was measured weekly. The data suggest that for surface mixes containing an
unmodified asphalt binder, rutting rates increase dramatically when the QC air voids fall
below 2.75%. At this point, removal and replacement of the HMA layer may be appropriate
for pavements expected to experience heavy traffic.
Mississippi
Mississippis original dense gravel mix placed in section S2 on the 2000 Track was designed
with 100 gyrations. Although rutting performance was exceptional, the section exhibited
extensive top-down cracking (Figure 7.1). Cracking was first noticed in this section after
approximately 12.5 million ESALs. Some of this cracking was not load related, as it was
observed in both the outside (research) and inside (non-trafficked) lane. A crack map for the
original section S2 near the end of the 2003 research cycle is shown in Figure 7.2. The
original 1.5 inch thick section S2 surface mix was milled to facilitate reconstruction for the
2006 Track. The milled material was stockpiled onsite and used as RAP in the replacement
mix, which was a similar dense-graded gravel mix designed using 85 gyrations. The reduced
gyration design mix in section S2 again exhibited very good rutting performance (3.0 mm);
however, at the very end of the 10 million ESAL traffic cycle extensive surface cracking was
noted. The final crack map for section S2 is shown in Figure 7.3
122
Transverse Offset (ft)
FIGURE 7.1 Cracking Severity in Section S2 after 30 Million ESALs.

12
11
10
9
8
7
6
5
4
3
2
1
0
25
35
45
55
65
75
85
95
105
115
125
135
145
155
165
175
Longitudinal Distance from Far Transverse Joint (ft)
FIGURE 7.2 Crack Map for Original Section S2 Near End of the 2003 Research Cycle.
123
Transverse Offset (ft)
12
11
10
9
8
7
6
5
4
3
2
1
0
25
35
45
55
65
75
85
95
105
115
125
135
145
155
165
175
Longitudinal Distance from Far Transverse Joint (ft)
FIGURE 7.3 Cracking Exhibited after 10 million ESALs in Section S2.

Following the successful performance of a 100 percent gravel SMA on the 2000 Track, the
Mississippi DOT chose to take gravel use one step further and evaluate a 100 percent gravel
OGFC on the 2006 Track. The new mix was placed at a target thickness of 1 inch. The new
performed well in terms of rutting (7.0 mm) and drainability. As seen in Figure 7.4, the
coefficient of permeability in section S3 was second only to section N13.
Missouri
Missouri DOT's "traditional" SMA specification required that 50 percent, by volume, of the
plus #8 material be a hard, durable aggregate. There is basically one source of material that
meets this requirement and it is located in Southeast Missouri. The limited availability of the
hard, durable aggregate resulted in a substantially higher cost for SMA. Therefore, SMA
was limited to MODOTs highest traffic roadways. These are known as commercial zones
and encompass the cities of Kansas City, St. Louis, Springfield and St. Joseph. Missouri's
objective on the 2003 Track was to determine if a 100 percent limestone SMA mix would
perform similar to the "traditional" SMA with the restrictive aggregate specification. Based
upon the findings from the 2003 cycle, Missouri now has a specification for 100 percent
limestone mixes and has expanded its use to all interstate and high volume routes. The
"traditional" SMA continues to be used inside the commercial zones, and the 100 percent
limestone mixes are used everywhere else. By the end of the 2009 construction season,
Missouri estimates it will have placed 176,000 tons of SMA using the revised aggregate
specification. Although the savings varies across the state, evidence suggests that on average
they are saving approximately $3.75 per ton. The revised specification has only been used
on 5 projects over two paving seasons, but Missouri estimates a total savings to date of
approximately $646,000. It is expected the revised specification will be used more in coming
paving seasons, leading to more savings for Missouri taxpayers (Dale Williams, unpublished
data).
124
30,000
Coefficient of Permeability (x 10-5 cm/sec)
25,000
20,000
15,000
10,000
5,000
0
N11
N12
N13
S3
S4
Permeable Surface Mix
FIGURE 7.4 Coefficient of Permeability in Drainable Surfaces Using Field

Permeameter.
North Carolina
As noted above, past Track research cycles have shown that fine mixes can provide rutting
performance that is comparable to coarse mixes. This finding was significant and led to many
states changing mix specification requirements to favor fine mixes in order to prevent
construction problems common with coarse mixes such as low mat densities and segregation.
The North Carolina DOT adjusted their Superpave specifications by relaxing their
specifications for Ninitial which opened the door to finer mixes (58). North Carolina has
continued traffic on two of its original coarse vs. fine comparison sections: S9 (coarse) and
S10 (fine). Now after 30 million ESALs and nine years, there is evidence from these
sections that fine-graded mixes are more resistant to top-down cracking than coarse mixes.
Section S10 shows less centerline cracking and less cracking from road scars (e.g., flat tires,
axle failures, etc.). Other comparison sections in the curves which remain from the 2000
research cycle, also support this observation.
Oklahoma
Previous Test Track experiments showed SMA mixtures performed well under accelerated
loading conditions. This knowledge was used by Oklahoma to gain confidence in specifying
and constructing SMA pavements. Oklahoma also used correlations between rutting
measured at the Track and APA test results to gain confidence in their new APA
specification (58).
The relationship between strain, temperature and speed was successfully quantified in
Oklahomas sections N8 (with a total asphalt thickness of 10 inches) and N9 (with a total
asphalt thickness of 14 inches). This outcome can be considered a successful validation of
M-E pavement analysis and design. Cracking was first noted in section N8 after
125
approximately 6.8 million ESALs. In consideration of beam fatigue performance in the
laboratory, strain measurements at the bottom of the pavement and cracking measurements
on the surface, it appears that section N9 can be considered a perpetual pavement. A
methodology is under development based on the results from these two sections, as well as
other results from all 3 research cycles, to relate laboratory beam fatigue results to multitemperature strains in the field.
South Carolina
At the end of the 2003 research cycle, absolute rut depths were measured at 4.1 mm in South
Carolinas section N13 (a 30% LA abrasion loss, granite SMA control mix) and 4.9 mm in
S1 (a 54% LA abrasion loss, SMA experimental mix). Although production problems with
the soft material in section S1 made it difficult to control gradations (resulting in a blend that
was slightly finer than desired), both sections had exhibited favorable rutting performance for
10 million ESALs. South Carolina decided to continue traffic on section S1 to quantify longterm performance of the mix with the softer aggregate while opting to discontinue research
on the N13 control mix. At the end of the 2006 research cycle (after the application of 20
million ESALs), absolute wire line rut depths in S1 averaged 7.7 mm (an additional 2.8 mm
of rutting over the second 10 million ESALs). In consideration of the extremely hot summer
of 2007 (average 7-day maximum air temperature = 100.5C), this mix performed well. It
was also observed that macrotexture increased in section S1 by approximately 0.15 mm
between 8 million ESALs (near the end of the 2003 Track) and 20 million ESALs (at the end
of the 2006 Track). In comparison, the 35% LA abrasion loss granite SMA placed by
Alabama in section W1 on the 2000 Track did not show any increase in macrotexture until
about 25 million ESALs (increasing by about 0.1 mm to 30 million ESALs). Macrotexture
measurements on both sections are shown in Figure 7.5 with no data shown for W1 before
the beginning of the 2003 research cycle in order to keep the sections on the same ESAL
scale. The difference in macrotexture performance between the two sections is presumably
due to the softer aggregate in section S1, and may be indicative of a difference in long term
durability.
126
1.5
1.4
1.3
1.2
Pavement Macrotexture (mm)
1.1
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0
5,000,000
10,000,000
15,000,000
20,000,000
Traffic Application (ESALs)

S1
W1 (beyond 10 million ESALs on 2000 Track)
FIGURE 7.5 Increase in Macrotexture of High LA Abrasion Loss SMA Surface.

Tennessee
Tennessee and Mississippi have designed test sections which relate aggregate selection to
pavement performance. Prior to the Test Track research, both states required blends of
limestone and crushed gravel for asphalt mixes. With their respective test sections, both
states have verified that 100% crushed gravel mixes can perform well in the field if proper
construction and design techniques are used (58).
Traffic continuation on Tennessees sections S4, S5 and E1 continued to yield valuable
performance information for experimental mixes. As seen previously in Figure 7.3, field
permeability measurements and observations indicated the OGFC in section S4 remained
drainable after 20 million ESALs with an average rut depth of only 6.5 mm. Likewise, the
75 gyration dense-graded gravel mix in section S5 exhibited only 3.4 mm of rutting; the
limestone SMA in section E1 exhibited 4.6 mm. Both of these measurements were taken
after the application of 20 million ESALs. The average rut depth of the new 65 gyration mix
in section S6 containing a blend similar to the older mix in S5 was only 1.5 mm. These
results have provided Tennessee with confidence in the lower gyration level design.
127
Texas
2006 was the first cycle that TXDOT participated in the NCAT Test Track. For this cycle
TXDOT was interested in evaluating a one inch rich bottom layer known as a CAM (crack
arresting mix). This mix was placed on section S12 after the Track foundation had been
carefully separated into slabs using a large diameter concrete masonry saw. A fine-graded
sand was used to fill the saw cuts to keep the slabs from healing back together during hot
weather. The CAM mix was a fine-graded 9.5 mm NMAS mix containing a PG70-22 binder
designed with 75 gyrations. A two inch lift of TXDOT Dense-Asphalt mix was placed over
the CAM mix. The Dense-Asphalt mix was designed with 50 gyrations and contained a
PG76-22. At the end of the cycle, no cracks were evident at the surface of the Texas test
section. However, the joints are perceptible at night under headlights. Additional coring is
being conducted to determine if any cracks currently exist in the CAM layer. 26.0 mm of
rutting was measured on the surface of the pavement by the end of the cycle. A coring
investigation revealed that all of the deformation was confined to the upper layer of densegraded mix. Very little deformation had occurred in the CAM lift.
128
CHAPTER 8 2009 TEST TRACK PLANS

Planning is currently underway for the 2009 NCAT Pavement Test Track (Phase IV). This
cycle of the Track is expected to consist of an even larger structural experiment as well as
more mill/inlay surface mixes, with research sponsorship expanded to include private sector
partners.
While individual test sections will still be optional on the 2009 Track, NCAT is also
developing a pre-designed six section Group Experiment that is intended to encompass
multiple timely issues that are important to the entire pavement community. All sections in
the Group Experiment will be supported by the same subgrade and base, and the total
thickness of all bituminous lifts will be 7 inches. This thickness was chosen because in past
studies, 7-inch sections exhibited significant performance differences within the planned
traffic cycle.
In addition to a control section that will be built with conventional HMA, two sections will
be built using different WMA technologies in every lift. Although the two WMA
technologies will be selected by the sponsors who choose to financially support the
experiment, it is envisioned that one of the sections will use a foaming process and the other
will be produced using an additive. These sections are proposed because reduced energy
demand, lower emissions, and enhanced workability make WMA technology a very
attractive alternative for the construction industry if it can be proven that early rutting,
moisture damage and structural performance are not compromised.
As a result of the rising cost of virgin materials, pavement engineers are also very interested
in high recycled content mixes. There are some concerns that the use of high percentages of
RAP in surface mixes may compromise durability. Likewise, there is concern that high RAP
content base and binder mixes may compromise fatigue resistance. It is critical that decision
makers determine whether high RAP content mixes are suitable for these applications so that
specification limits can be set at the highest level that exhibits performance characteristics
comparable to virgin mixes. In order to address this issue, one section will be built with a
high RAP content in lower lift(s) and low RAP contents in upper lift(s). Another section will
be built with high RAP contents in both lower and upper lifts.
Many state DOTs are using drainable surface mixes in order to improve wet weather driving
visibility, reduce accident/fatality rates, and reduce noise created by pavement tire
interaction. Although drainable surface mixes have aggregate structures that are very similar
to rut resistant SMA mixes, it is typically assumed they do not contribute to the load carrying
potential of the pavement structure. The sixth section in the Group Experiment will be
built identical to the control section, except that the conventional surface mix will be
replaced with a drainable surface mix.
By monitoring response instrumentation (i.e., pressure plates and strain gauges) installed in
each of these sections at the time they are constructed and by documenting changing surface
conditions (rutting, roughness, cracking, etc.) under heavy truck traffic, it will be possible to
129
compare both surface and structural performance. It is expected that this information will
provide for the optimization of specifications regarding the deployment of these modern
technologies on the pavement infrastructure with a high level of confidence.
Utilization of as many sections as possible for structural purposes would facilitate the
implementation of M-E methods for structural pavement design. For example, the
development of the new MEPDG represents a significant change and advancement over
existing design methodologies. Historically, the structural design of asphalt pavements has
been largely empirical based upon vehicle designs, axle loads, and material properties. The
new design guide, however, relies heavily on principles of engineering mechanics to produce
thickness designs that control specific modes of pavement distress. Before this new
methodology gains wide acceptance or use, it must be validated and calibrated to ensure that
it provides adequate design guidance using modern methods and materials under traffic by
actual design vehicles. Calibration of the conservative distress models that could eliminate
only a 10 percent margin of error in excess design thickness would generate an annual
taxpayer savings nationwide of as much as one billion dollars. To this end, there is a need
for a full-scale structural experiment to validate the methodology.
130
REFERENCES
1. Brown, E.R., L.A. Cooley, D. Hanson, C. Lynn, B. Powell, B. Prowell, and D.
Watson, NCAT Test Track Design, Construction, and Performance, NCAT 02-12,
National Center for Asphalt Technology, Auburn University, 2002.
2. Timm, D.H. Design, Construction, and Instrumentation of the 2006 Test Track
Structural Study, NCAT Draft Report, Auburn University, 2008.
3. Wang, J. et al, Windows-Based Top-Down Cracking Design Tool for Florida: Using
Energy Ratio Concept, In Transportation Research Record: Journal of the
Transportation Research Board, No. 2037, Transportation Research Board of the
National Academies, Washington, D.C., 2007, pp 86-96.
4. Roque, R. et al, Development and Field Evaluation of Energy-Based Criteria for
Top-Down Cracking Performance of Hot Mix Asphalt, Journal of the Association of
Asphalt Paving Technologists, Vol. 73, 2004, pp. 229-260.
5. Zhang, Z. et al, Evaluation of Laboratory-Measured Crack Growth Rate for Asphalt
Mixtures, In Transportation Research Record: Journal of the Transportation
Research Board, No. 1767, Transportation Research Board of the National
Academies, Washington, D.C., 2001a, pp. 67-75.
6. Zhang, Z. et al, Identification and Verification of a Suitable Crack Growth Law
Journal of the Association of Asphalt Paving Technologists, Vol. 70, 2001b, pp. 816827.
7. Birgisson, B. et al, Hot Mix Asphalt Fracture Mechanics: A Fundamental Crack
Growth Law for Asphalt Mixtures, Journal of the Association of Asphalt Paving
Technologists, Vol. 71, 2002, pp. 816-827.
8. Roque, R. et al, Implementation of SHRP Indirect Tension Tester to Mitigate
Cracking in Asphalt Pavements and Overlays. Final Report for FDOT BA-546
Contract, University of Florida, Gainesville, FL, 2002.
9. Timm, D.H. et al, Forensic Investigation and Validation of Energy Ratio Concept,
Submitted to the 88th Annual Meeting of the Transportation Research Board.
Transportation Research Record (in press), 2008.
10. Willis, J.R. and D.H. Timm. Forensic Investigation of a Rich-Bottom Pavement.
Report No. 06-04, National Center for Asphalt Technology, Auburn University, 2006.
11. Willis, J.R. and D.H. Timm. A Forensic Investigation of Debonding in a RichBottom Pavement, In Transportation Research Record: Journal of the Transportation
Research Board, No. 2040, Transportation Research Board of the National
Academies, Washington, D.C., 2007, pp. 107-114.
12. Federal Highway Administration RAP Expert Task Group, 10 Obstacles for
Increased RAP Use, HMA Recycling Expert Task Group website:
http:/www.ncat.us/RAP/Past%20RAP%20ETGs/05-07.html , accessed November 14,
2008.
13. West, R. et al, Laboratory and Accelerated Field Performance Testing of Moderate
and High RAP Content Mixes at the NCAT Test Track, Submitted to the 88th
Annual Meeting of the Transportation Research Board, (in review), 2009.
14. Smit, A.d.F and B. Waller, Sound Pressure And Intensity Evaluations Of Low Noise
Pavement Structures With Open Graded Asphalt Mixtures, NCAT Report 07-02,
National Center for Asphalt Technology at Auburn University, Auburn, AL, 2007.
131
15. Smit, A.d.F and B. Waller, Sound Pressure and Intensity Evaluations of Low Noise
Pavement Structures with Dense-Graded and Stone Matrix Asphalt Mixtures, Draft
NCAT Report 07-03, National Center for Asphalt Technology at Auburn University,
Auburn, AL, 2007.
16. Smit, A.d.F and B. Waller, Evaluation Of The Ultra-Light Inertial Profiler (ULIP)
For Measuring Surface Texture Of Pavements, NCAT Report 07-01, National
Center for Asphalt Technology at Auburn University, Auburn, AL, 2007.
17. Powell, B. Construction and Performance of Georgias Permeable Surface Mixes on
the 2006 NCAT Pavement Test Track, Submitted to the 88th Annual Meeting of the
Transportation Research Board. Transportation Research Record (in press), 2009.
18. Beuving, E. European Asphalt Pavement Association, Presentation made at the
National Center for Asphalt Technology, March 2006.
19. Ingram, L. S., Herbold, K. D., Baker, T. E., Brumfield, J. W., Felag, M. E., Ferragut,
T. R., Grogg, M. G., Lineman, L. R., and R. O. Rasmussen, Superior Materials,
Advanced Test Methods, and Specifications in Europe, FHWAPL-04-007, Federal
Highway Administration International Technology Exchange Program, April 2004.
20. Asphalt Institute. Superpave Mix Design (SP-2). Lexington, KY, 1996.
21. National Asphalt Pavement Association. An Industry Discussion on Superpave
Implementation. Lanham, MD.
22. Kandhal, P. and Cooley L.A., Jr. Accelerated Laboratory Rutting Tests: Evaluation of
the Asphalt Pavement Analyzer. NCHRP report 508, Transportation Research Board,
Washington, D.C., 2008.
23. Witczak, M. Simple Performance Tests: Summary of Recommended Methods and
Database. NCHRP Report 547, NCHRP, TRB, National Research Council,
Washington D.C., 2005.
24. Eres Consultants Division (2004), Guide For Mechanistic-Empirical Pavement
Design of New and Rehabilitated Pavement Structures, Final Report, NCHRP 1-37A.
25. Bonaquist, R., D. Christensen, and W. Stump. Simple Performance Tester for
Superpave Mix Design: First Article Development and Evaluation. NCHRP report
513, NCHRP, TRB, National Research Council, Washington D.C., 2003.
26. NCAT Pavement Test Track. www.pavetrack.com, accessed July 27, 2008.
27. Taylor, A.J. Mechanistic Characterization of Resilient Moduli for Unbound
Pavement Layer Materials. M.S. Thesis. Auburn University. Auburn, AL. August
2008.
28. Robbins, M.M. and D.H. Timm, Temperature and Velocity Effects on a Flexible
Perpetual Pavement Transportation Research Record Preprint, 2009.
29. Willis, J.R., Field-Based Strain Thresholds for Flexible Perpetual Pavement
Design, PhD Dissertation, Auburn University, 2009.
30. Eres Consultants Division, Guide For Mechanistic-Empirical Pavement Design of
New and Rehabilitated Pavement Structures; Appendix CC-3, Updated Traffic
Frequency Calculation for Asphalt Layers, Final Document, NCHRP 1-37A, 2003.
31. Newcomb, D.E., M. Buncher, and I.J. Huddleston. Concepts of Perpetual Pavements.
In Transportation Research Circular: Journal of the Transportation Research Board,
No. 503, TRB, National Research Council, Washington, D.C., 2001, pp 4-11.
32. Newcomb, D.E. and H. Von Quintus. Wanted: Transfer Functions Experience
Needed. Hot Mix Asphalt Technology, Vol. 7, No. 8, 2002, pp 22-25.
132
33. Priest, A.L. and D.H. Timm. Methodology and Calibration of Fatigue Transfer
Functions for Mechanistic-Empirical Flexible Pavement Design, Report No. 06-03,
National Center for Asphalt Technology, Auburn University, 2006.
34. Timm, D.H. and A.L. Priest, Material Properties of the 2003 NCAT Test Track
Structural Study, Report No. 06-01, National Center for Asphalt Technology,
Auburn University, 2006.
35. Miner, M.A. Estimation of Fatigue Life with Emphasis on Cumulative Damage.
Metal Fatigue, edited by Sines and Wiseman, McGraw Hill, 1959, pp 278-89.
36. St. Martin, J., J.T. Harvey, F. Long, E. Lee, C.L. Monismith, and K. Herritt. LongLife Rehabilitation Design and Construction, Transportation Research Circular,
Number 503, 2001, pp 50-65.
37. Walubita, L.F., W. Liu, T. Scullion, and J. Leidy. Modeling Perpetual Pavements
Using the Flexible Pavement System (FPS) Software, Transportation Research
Board 2008 Annual Meeting, CD-ROM.
38. Al-Qadi, I.L., H. Wang, P.J. Yoo, and S. H. Dessouky. Dynamic Analysis and Insitu Validation of Perpetual Pavement Response to Vehicular Loading,
Transportation Research Board 2008 Annual Meeting, CD-ROM.
39. Willis, J.R. A Synthesis of Practical and Appropriate Instrumentation Use for
Accelerated Pavement Testing, International Conference on Accelerated Pavement
Testing, Madrid, Spain, 2008.
40. Hornyak, N.J., J.A. Crovetti, D.E. Newman, and J.P. Schabelski. Perpetual
Pavement Instrumentation for the Marquette Interchange Project Phase 1, SPR
#0092-06-01, Wisconsin Highway Research Program, August 2007.
41. Romanoschi, S.A., A.J. Gisi, M. Portillo, and C. Dumitru. The first findings from
the Kansas Perpetual Pavements experiment Transportation Research Board 2008
Annual Meeting, CD-ROM.
42. Yang, Y. et al, Perpetual Pavement Design in China, International Conference on
Perpetual Pavement, Ohio Research Institute for Transportation and the Environment,
2005.
43. American Association of State Highway and Transportation Officials. AASHTO
Guide for Design of Pavement Structures. Washington, D.C., 1993.
44. Willis, J.R. and D.H. Timm, Repeatability of Asphalt Strain Measurements under
Full Scale Dynamic Loading, Transportation Research Board Preprint, 2008.
45. Selvaraj, S.I., Development of Flexible Pavement Rut Prediction Models from the
NCAT Test Track Structural Study Sections Data, PhD Dissertation, Auburn
University, 2007.
46. Priest, A.L. and D.H. Timm, Mechanistic Comparison of Wide-Base Single Versus
Standard Dual Tire Configurations, Transportation Research Record No. 1949,
Transportation Research Board, 2006b, pp. 155-163
47. Seeds, S.B., S.H. Alavi, W.C. Ott, M. Mikhail, and J.A. Mactutis. Evaluation of
Laboratory Determined and Nondestructive Test Based Resilient Modulus Values
from the WesTrack Experiment. Nondestructive Testing and Backcalculation of
Moduli: Third Volume. ASTM STP 1375. S.D. Tayabji and E.O. Lukanen, Eds.,
American Society of Testing and Materials, West Conshohocken, PA, 2000.
48. Parker, F. and D.J. Elton. Methods for Evaluating Resilient Moduli of Paving
Materials. Auburn University Highway Research Center: Auburn, AL. 1990.
133
49. American Association of State Highway and Transportation Officials T 315-06.
Standard Test Method for Determining the Rheological Properties of Asphalt Binder
Using Dynamic Shear Rheometer (DSR). AASHTO, 2008.
50. ASTM D2983-04a(2004). Standard Test Method for Low-Temperature Viscosity of
Lubricants Measured by Brookfield Viscometer. ASTM International, West
Conshohocken, Pa., 2004. www.astm.org.
51. ARA Inc., Eres Division. Part 2 Design Inputs: Chapter 2 Material Characterization.
Guide For Mechanistic-Empirical Pavement Design of New and Rehabilitated
Pavement Structures, Final Document, NCHRP 1-37A, 2004.
52. Bari, J. and M.W. Witczak. Development of a New Revised Version of the Witczak
E* Predictive Model for Hot Mix Asphalt Mixtures. Proceedings of the Association
of Asphalt Paving Technologists, Vol. 75, 2006, pp381-423.
53. Andrei, D., M.W. Witczak, and W. Mirza. Appendix CC-4: Development of a
Revised Predictive Model for the Dynamic (Complex) Modulus of Asphalt Mixtures.
Development of the 2002 Guide for the Design of New and Rehabilitated Pavement
Structures, Final Document, NCHRP 1-37A, 1999.
54. Christensen, Jr., D.W., T. Pellinen, and R.F. Bonaquist. Hirsch Model for Estimating
the Modulus of Asphalt Concrete. Proceedings of the Association of Asphalt Paving
Technologists, Vol. 72, 2003, pp97-121.
55. Robbins, M.M. Masters Thesis. Auburn University. Draft Copy, 2009.
56. Azari, H., G. Al-Khateeb, A, Shenoy, and N. Gibson. Comparison of Measured SPT
E* of ALF Mixtures with Predicted E* Using NCHRP 1-37A and Witczaks New
Equations. In Proceedings of the 86th Annual Transportation Research Board, TRB,
National Research Council, Washington, D.C., 2007.
57. Prowell, B., E.R. Brown, J. Daniel, S. Bhattacharjee, H. Von Quintus, S. Carpenter,
S. Shen, M. Anderson, A.K. Swamy, and S. Maghsoodloo. Endurance Limits of Hot
Mix Asphalt Mixtures to Prevent Fatigue Cracking in Flexible Pavements. NCHRP
Report 9-38, Updated Final Draft Report, National Center for Asphalt Technology,
Auburn, AL. May 2008.
58. Brown, E.R., B. Powell, R. West, and D. Timm. NCAT Test Track Findings, NCAT
Report, National Center for Asphalt Technology, Auburn University, 2005.
134

09-08 Accelerated Performance Testing On The 2006 NCAT Pavement Test Track

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

09-08 Accelerated Performance Testing On The 2006 NCAT Pavement Test Track

Uploaded by

Copyright:

Available Formats

ACCELERATED PERFORMANCE TESTING ON

THE 2006 NCAT TEST TRACK

ACCELERATED PERFORMANCE TESTING ON THE 2006

National Center for Asphalt Technology

LIST OF ACRONYMS ............................................................................................................ 1

FIGURE 1.1 The NCAT Test Track.

FIGURE 1.2 Layout of Test Track (1).

FIGURE 2.1 2006 NCAT Test Track.

Depth From Pavement Surface, in.

FIGURE 3.1 FDOT Test Sections (9).

N1 milled and inlaid

FIGURE 3.2 Measured Strain and Temperature versus Date (9).

Longitudinal Microstrain Under Single Axle

Rut Depth (mm)

Rut Depth @ 8,000 Cycles (mm)

45% RAP Virgin PG

FIGURE 3.4 Rut Depths from APA Testing (13).

Fatigue Life (1000 cycles)

45% RAP 45% RAP

FIGURE 3.8 Summary of Global Sound Intensity Levels (15).

FIGURE 3.9 RMS Filter Applied to ULIP Surface Profile (15).

Long-term friction loss trend

FIGURE 3.10 Generic HMA Surface Friction Performance Curve.

FIGURE 3.11 Aggregate Gradations of the Sections.

GDOT Permeable Test Section

FIGURE 3.12 Post-Construction Smoothness Comparison (14).

In-Place Air Voids

GDOT Permeable Sublot

Bond Strength (MPa)

FIGURE 3.15 Laboratory Permeability Measurements from Cut Cores (14).

International Roughness Index (m/km)

Equivalent Single Axle Loadings

FIGURE 3.16 Roughness Measurements as a Function of Truck Traffic (14).

Mean Texture Depth (mm)

Equivalent Single Axle Loadings

FIGURE 3.17 Macrotexture Measurements as a Function of Truck Traffic (14).

FIGURE 3.18 SPL of the Sections over Time.

FIGURE 3.19 Noise Spectra Dec-06 Measurements.

FIGURE 3.20 Noise Spectra Mar-07 Measurements.

FIGURE 3.21 Noise Spectra May-07 Measurements.

FIGURE 3.22 Noise Spectra Dec-07 Measurements.

FIGURE 3.23 Noise Spectra Mar-08 Measurements.

FIGURE 3.24 Noise Spectra Jan-09 Measurements.

Binder RAP% NMAS

Manual Auto Manual Auto Power Francken

Field Rut Depth (mm)

10 mil ESAL Rut

Field Rut Depth (mm)

Fn Acceptance Criterion = 800 cycles

r = - 0.879, p-value = 0.004

Field Rut Depth (m m )

Avg Rut Depth (mm)

Equivalent Single Axle Loadings

FIGURE 3.29 Rutting Progression in Section S8B Showing Replacement Mix.

FIGURE 3.30 Relationship Between Rutting and QC Air Voids.

Alabama & FHWA

PG 64-22 (2% Air Voids)

FIGURE 4.1 As-Built Pavement Cross Sections (2).

Asphalt Strain Gauge

FIGURE 4.2 N9 Supplemental Instrumentation (2).

Strain vs. Temperature

Application of Regression Models

FIGURE 5.1 Predicted Strain by Temperature for a Single Axle.

10465 .42 14.7